Académique Documents
Professionnel Documents
Culture Documents
This product or document is protected by copyright and distributed under licenses restricting its use, copying, distribution, and decompilation. No
part of this product or document may be reproduced in any form by any means without prior written authorization of Sun and its licensors, if any.
Third-party software, including font technology, is copyrighted and licensed from Sun suppliers.
Parts of the product may be derived from Berkeley BSD systems, licensed from the University of California. UNIX is a registered trademark in the U.S.
and other countries, exclusively licensed through X/Open Company, Ltd.
Sun, Sun Microsystems, the Sun logo, docs.sun.com, AnswerBook, AnswerBook2, and Solaris are trademarks, registered trademarks, or service marks
of Sun Microsystems, Inc. in the U.S. and other countries. All SPARC trademarks are used under license and are trademarks or registered trademarks
of SPARC International, Inc. in the U.S. and other countries. Products bearing SPARC trademarks are based upon an architecture developed by Sun
Microsystems, Inc.
The OPEN LOOK and Sun Graphical User Interface was developed by Sun Microsystems, Inc. for its users and licensees. Sun acknowledges the
pioneering efforts of Xerox in researching and developing the concept of visual or graphical user interfaces for the computer industry. Sun holds a
non-exclusive license from Xerox to the Xerox Graphical User Interface, which license also covers Suns licensees who implement OPEN LOOK GUIs
and otherwise comply with Suns written license agreements.
Federal Acquisitions: Commercial SoftwareGovernment Users Subject to Standard License Terms and Conditions.
DOCUMENTATION IS PROVIDED AS IS AND ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES,
INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT, ARE
DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID.
Copyright 2004 Sun Microsystems, Inc. 4150 Network Circle, Santa Clara, CA 95054 U.S.A. Tous droits rservs.
Ce produit ou document est protg par un copyright et distribu avec des licences qui en restreignent lutilisation, la copie, la distribution, et la
dcompilation. Aucune partie de ce produit ou document ne peut tre reproduite sous aucune forme, par quelque moyen que ce soit, sans
lautorisation pralable et crite de Sun et de ses bailleurs de licence, sil y en a. Le logiciel dtenu par des tiers, et qui comprend la technologie relative
aux polices de caractres, est protg par un copyright et licenci par des fournisseurs de Sun.
Des parties de ce produit pourront tre drives du systme Berkeley BSD licencis par lUniversit de Californie. UNIX est une marque dpose aux
Etats-Unis et dans dautres pays et licencie exclusivement par X/Open Company, Ltd.
Sun, Sun Microsystems, le logo Sun, docs.sun.com, AnswerBook, AnswerBook2, et Solaris sont des marques de fabrique ou des marques dposes, ou
marques de service, de Sun Microsystems, Inc. aux Etats-Unis et dans dautres pays. Toutes les marques SPARC sont utilises sous licence et sont des
marques de fabrique ou des marques dposes de SPARC International, Inc. aux Etats-Unis et dans dautres pays. Les produits portant les marques
SPARC sont bass sur une architecture dveloppe par Sun Microsystems, Inc.
Linterface dutilisation graphique OPEN LOOK et Sun a t dveloppe par Sun Microsystems, Inc. pour ses utilisateurs et licencis. Sun reconnat
les efforts de pionniers de Xerox pour la recherche et le dveloppement du concept des interfaces dutilisation visuelle ou graphique pour lindustrie
de linformatique. Sun dtient une licence non exclusive de Xerox sur linterface dutilisation graphique Xerox, cette licence couvrant galement les
licencis de Sun qui mettent en place linterface dutilisation graphique OPEN LOOK et qui en outre se conforment aux licences crites de Sun.
CETTE PUBLICATION EST FOURNIE EN LETAT ET AUCUNE GARANTIE, EXPRESSE OU IMPLICITE, NEST ACCORDEE, Y COMPRIS DES
GARANTIES CONCERNANT LA VALEUR MARCHANDE, LAPTITUDE DE LA PUBLICATION A REPONDRE A UNE UTILISATION
PARTICULIERE, OU LE FAIT QUELLE NE SOIT PAS CONTREFAISANTE DE PRODUIT DE TIERS. CE DENI DE GARANTIE NE
SAPPLIQUERAIT PAS, DANS LA MESURE OU IL SERAIT TENU JURIDIQUEMENT NUL ET NON AVENU.
031111@6671
Contents
Preface 15
2 Link-Editor 25
Invoking the Link-Editor 26
Direct Invocation 26
Using a Compiler Driver 27
Specifying the Link-Editor Options 27
Input File Processing 28
Archive Processing 29
Shared Object Processing 30
Linking With Additional Libraries 31
Initialization and Termination Sections 36
Symbol Processing 38
Symbol Resolution 38
Undefined Symbols 42
3
Tentative Symbol Order Within the Output File 46
Defining Additional Symbols 46
Reducing Symbol Scope 52
External Bindings 56
String Table Compression 56
Generating the Output File 57
Relocation Processing 58
Displacement Relocations 58
Debugging Aids 60
3 Runtime Linker 63
Shared Object Dependencies 64
Locating Shared Object Dependencies 64
Directories Searched by the Runtime Linker 64
Configuring the Default Search Paths 67
Dynamic String Tokens 67
Relocation Processing 67
Symbol Lookup 68
When Relocations Are Performed 71
Relocation Errors 72
Loading Additional Objects 73
Lazy Loading of Dynamic Dependencies 74
Initialization and Termination Routines 76
Initialization and Termination Order 77
Security 80
Runtime Linking Programming Interface 81
Loading Additional Objects 82
Relocation Processing 83
Obtaining New Symbols 89
Feature Checking 92
Debugging Aids 93
Debugging Library 93
Debugger Module 96
Contents 5
Invoking the Auditing Interface 157
Recording Local Auditors 157
Audit Interface Functions 158
Audit Interface Example 162
Audit Interface Demonstrations 163
Audit Interface Limitations 163
Runtime Linker Debugger Interface 164
Interaction Between Controlling and Target Process 165
Debugger Interface Agents 166
Debugger Exported Interface 167
Debugger Import Interface 175
Contents 7
Versioning an Existing (Non-versioned) Shared Object 314
Updating a Versioned Shared Object 315
Adding New Symbols 315
Internal Implementation Changes 316
New Symbols and Internal Implementation Changes 316
Migrating Symbols to a Standard Interface 317
Index 335
9
TABLE 726 SPARC: ELF Symbol Table Entry: Register Symbol 229
TABLE 727 SPARC: ELF Register Numbers 229
TABLE 728 ELF si_boundto Reserved Values 230
TABLE 729 ELF Syminfo Flags 230
TABLE 730 ELF Version Definition Structure Versions 232
TABLE 731 ELF Version Definition Section Flags 232
TABLE 732 ELF Version Dependency Indexes 233
TABLE 733 ELF Version Dependency Structure Versions 235
TABLE 734 ELF Version Dependency Structure Flags 235
TABLE 735 ELF Segment Types 238
TABLE 736 ELF Segment Flags 240
TABLE 737 ELF Segment Permissions 240
TABLE 738 SPARC: ELF Program Header Segments (64K alignment) 243
TABLE 739 x86: ELF Program Header Segments (64K alignment) 244
TABLE 740 SPARC: ELF Example Shared Object Segment Addresses 247
TABLE 741 x86: ELF Example Shared Object Segment Addresses 247
TABLE 742 ELF Dynamic Array Tags 249
TABLE 743 ELF Dynamic Flags, DT_FLAGS 256
TABLE 744 ELF Dynamic Flags, DT_FLAGS_1 257
TABLE 745 ELF Dynamic Position Flags, DT_POSFLAG_1 259
TABLE 746 ELF Dynamic Feature Flags, DT_FEATURE_1 260
TABLE 747 SPARC: Procedure Linkage Table Example 262
TABLE 748 64-bit SPARC: Procedure Linkage Table Example 265
TABLE 749 x86: Absolute Procedure Linkage Table Example 268
TABLE 750 x86: Position-Independent Procedure Linkage Table Example 268
TABLE 81 ELF PT_TLS Program Header Entry 273
TABLE 82 SPARC: 32-bit and 64-bit General Dynamic Thread-Local Variable
Access Codes 279
TABLE 83 SPARC: 32-bit and 64-bit Local Dynamic Thread-Local Variable Access
Codes 280
TABLE 84 SPARC: 32-bit Initial Executable Thread-Local Variable Access Codes
282
TABLE 85 SPARC: 64-bit Initial Executable Thread-Local Variable Access Codes
283
TABLE 86 SPARC: 32-bit and 64-bit Local Executable Thread-Local Variable Access
Codes 283
TABLE 87 SPARC: Thread-Local Storage Relocation Types 284
TABLE 88 x86: General Dynamic Thread-Local Variable Access Codes 286
TABLE 89 x86: Local Dynamic Thread-Local Variable Access Codes 287
Tables 11
12 Linker and Libraries Guide April 2004
Figures
13
14 Linker and Libraries Guide April 2004
Preface
Intended Audience
This manual is intended for a range of programmers who are interested in the Solaris
linkers, from the curious beginner to the advanced user.
Beginners learn the principle operations of the link-editor and runtime linker.
Intermediate programmers learn to create, and use, efficient custom libraries.
Advanced programmers, such as language-tools developers, learn how to interpret
and generate object files.
Not many programmers should need to read this manual from cover to cover.
15
Organization
Chapter 1 gives an overview of the linking processes under the Solaris operating
environment, together with an introduction of new features added with this release.
This chapter is intended for all programmers.
Chapter 2 describes the functions of the link-editor, its two modes of linking (static and
dynamic), scope and forms of input, and forms of output. This chapter is intended for
all programmers.
Chapter 7 is a reference chapter on ELF files. This chapter is intended for advanced
programmers.
Chapter 9 describes the mapfile directives to the link-editor, which specify the layout
of the output file. This chapter is intended for advanced programmers.
Appendix D provides an overview of new features and updates that have been added
to the link-editors and indicates to which release they were added.
Throughout this document, all command-line examples use sh(1) syntax, and all
programming examples are written in the C language.
Typographic Conventions
The following table describes the typographic changes used in this book.
AaBbCc123 The names of commands, files, and Edit your .login file.
directories; on-screen computer output
Use ls -a to list all files.
machine_name% you have
mail.
Preface 17
TABLE P1 Typographic Conventions (Continued)
Typeface or Symbol Meaning Example
AaBbCc123 Book titles, new words, or terms, or Read Chapter 6 in Users Guide.
words to be emphasized.
These are called class options.
You must be root to do this.
Shell Prompt
This manual describes the operations of the Solaris link-editor and runtime linker,
together with the objects on which they operate. The basic operation of the Solaris
linkers involves the combination of objects. The symbolic references from one object
are connected to the symbolic definitions within another object. This operation is often
referred to as binding.
These areas, although separable into individual topics, have a great deal of overlap.
While explaining each area, this document brings together the connecting principles.
19
Link-Editing
Link-editing takes a variety of input files, typically generated from compilers,
assemblers, or ld(1). The link-editor concatenates and interprets the data within these
input files to form a single output file. Although the link-editor provides numerous
options, the output file that it produces is one of four basic types:
Relocatable object A concatenation of input relocatable objects that can be used in
subsequent link-edit phases.
Static executable A concatenation of input relocatable objects that has all symbolic
references bound to the executable, and thus represents a ready-to-run process.
Dynamic executable A concatenation of input relocatable objects that requires
intervention by the runtime linker to produce a runnable process. A dynamic
executable might still need symbolic references bound at runtime, and can have
one or more dependencies in the form of shared objects.
Shared object A concatenation of input relocatable objects that provides services
that might be bound to a dynamic executable at runtime. The shared object can
have dependencies on other shared objects.
These output files, and the key link-editor options used to create them, are shown in
Figure 11.
Dynamic executables and shared objects are often referred to jointly as dynamic objects.
Dynamic objects are the main focus of this document.
ld
-dn -dy
-r -G
During process execution the facilities of the runtime linker are made available, and
can be used to extend the process address space by adding additional shared objects
on demand. The two most common components involved in runtime linking are
dynamic executables and shared objects.
Dynamic executables are applications that are executed under the control of a runtime
linker. These applications usually have dependencies in the form of shared objects,
which are located and bound by the runtime linker to create a runnable process.
Dynamic executables are the default output file generated by the link-editor.
Shared objects provide the key building block to a dynamically linked system. A
shared object is similar to a dynamic executable; however, shared objects have not yet
been assigned a virtual address.
Dynamic executables usually have dependencies on one or more shared objects. That
is, the shared object(s) must be bound to the dynamic executable to produce a
runnable process. Because shared objects can be used by many applications, aspects of
their construction directly affect shareability, versioning, and performance.
You can distinguish the processing of shared objects by either the link-editor or the
runtime linker by referring to the environments in which the shared objects are being
used:
compilation environment
Shared objects are processed by the link-editor to generate dynamic executables or
other shared objects. The shared objects become dependencies of the output file
being generated.
runtime environment
Shared objects are processed by the runtime linker, together with a dynamic
executable, to produce a runnable process.
Dynamic Linking
Dynamic linking is a term often used to embrace those portions of the link-editing
process that generate dynamic executables and shared objects, together with the
runtime linking of these objects to generate a runnable process. Dynamic linking
enables multiple applications to use the code provided by a shared object by enabling
the application to bind to the shared object at runtime.
The Solaris ABI is a technological descendent for work on ABIs that started with the
System V Application Binary Interface and the successor work performed by SPARC
International, Inc. for SPARC processors called the SPARC Compliance Definition
(SCD).
The operations of the link-editors on 32bit and 64bit objects is identical. This
document typically uses 32bit examples. Cases where 64bit processing differs from
the 32bit processing are highlighted.
For more information regarding 64bit applications, refer to the Solaris 64-bit
Developers Guide.
Environment Variables
The link-editors support a number of environment variables that begin with the
characters LD_, for example LD_LIBRARY_PATH. Each environment variable can exist
in its generic form, or can be specified with a _32 or _64 suffix, for example
LD_LIBRARY_PATH_64. This suffix makes the environment variable specific,
respectively, to 32-bit or 64bit processes. This suffix also overrides any generic,
non-suffixed, version of the environment variable that may be in effect.
Support Tools
The Solaris operating environment also provides several support tools and libraries.
These tools provide for the analysis and inspection of these objects and the linking
processes. These tools include elfdump(1), nm(1), dump(1), ldd(1), pvs(1), elf(3ELF),
and a linker debugging support library. Throughout this document, many discussions
are augmented with examples of these tools.
Link-Editor
The link-editing process creates an output file from one or more input files. The
creation of the output file is directed by the options supplied to the link-editor
together with the input sections provided by the input files.
All files are represented in the executable and linking format (ELF). For a complete
description of the ELF format see Chapter 7. For this introduction, however, it is first
necessary to introduce two ELF structures, sections and segments.
Sections are the smallest indivisible units that can be processed within an ELF file.
Segments are a collection of sections that represent the smallest individual units that
can be mapped to a memory image by exec(2) or by the runtime linker ld.so.1(1).
Although there are many types of ELF sections, they all fall into two categories with
respect to the link-editing phase:
Sections that contain program data, whose interpretation is meaningful only to the
application itself, such as the program instructions .text and the associated data
.data and .bss.
Sections that contain link-editing information, such as the symbol table information
found from .symtab and .strtab, and relocation information such as
.rela.text.
Basically, the link-editor concatenates the program data sections into the output file. The
link-editing information sections are interpreted by the link-editor to modify other
sections or to generate new output information sections used in later processing of the
output file.
25
segments.
It reads symbol table information from both relocatable objects and shared objects
to verify and unite references with definitions, and usually generates a new symbol
table, or tables, within the output file.
It reads relocation information from the input relocatable objects and applies this
information to the output file by updating other input sections. In addition, output
relocation sections might be generated for use by the runtime linker.
It generates program headers that describe all segments created.
It generates dynamic linking information sections if necessary, which provide
information such as shared object dependencies and symbol bindings to the
runtime linker.
The process of concatenating like sections and associating sections to segments is carried
out using default information within the link-editor. The default section and segment
handling provided by the link-editor is usually sufficient for most link-edits. However,
these defaults can be manipulated using the -M option with an associated mapfile.
See Chapter 9.
Direct Invocation
When you invoke the link-editor directly, you have to supply every object file and
library required to create the intended output. The link-editor makes no assumptions
about the object modules or libraries that you meant to use in creating the output. For
example, when you issue the command:
$ ld test.o
the link-editor creates a dynamic executable named a.out using only the input file
test.o. For the a.out to be a useful executable, it should include startup and exit
processing code. This code can be language or operating system specific, and is
usually provided through files supplied by the compiler drivers.
When creating runtime objects such as executables and shared objects, you should use
a compiler driver to invoke the link-editor. Invoking the link-editor directly is
recommended only when creating intermediate relocatable objects when using the -r
option.
Note The actual files included by your compiler driver and the mechanism used to
display the link-editor invocation might differ.
The -R and -L options are interpreted by the link-editor and prepended to any
command-line options received from the compiler driver.
Chapter 2 Link-Editor 27
The link-editor parses the entire option list for any invalid options or any options with
invalid associated arguments. When either of these cases is found, a suitable error
message is generated. If the error is deemed fatal, the link-edit terminates. In the
following example, the illegal option -X is identified, and the illegal argument to the
-z option is caught by the link-editors checking.
$ ld -X -z sillydefs main.o
ld: illegal option -- X
ld: fatal: option -z has illegal argument sillydefs
The link-editor also checks the option list for any fatal inconsistencies. For example:
$ ld -dy -a main.o
ld: fatal: option -dy and -a are incompatible
After processing all options, if no fatal error conditions have been detected, the
link-editor proceeds to process the input files.
See Appendix A for the most commonly used link-editor options, and the ld(1) man
page for a complete description of all link-editor options.
Under static mode, the link-editor accepts only relocatable objects or archive libraries
as input files. Under dynamic mode, the link-editor also accepts shared objects.
Relocatable objects represent the most basic input file type to the link-editing process.
The program data sections within these files are concatenated into the output file image
being generated. The link-edit information sections are organized for later use, but do
not become part of the output file image, as new sections are generated to take their
places. Symbols are gathered into an internal symbol table for verification and
resolution. This table is then used to create one or more symbol tables in the output
image.
Archive Processing
Archives are built using ar(1), and usually consist of a collection of relocatable objects
with an archive symbol table. This symbol table provides an association of symbol
definitions with the objects that supply these definitions. By default, the link-editor
provides selective extraction of archive members. When the link-editor reads an
archive, it uses information within the internal symbol table it is creating to select only
the objects from the archive it requires to complete the binding process. You can also
explicitly extract all members of an archive.
Under selective archive extraction, a weak symbol reference does not extract an object
from an archive unless the -z weakextract option is in effect. See Simple
Resolutions on page 39 for more information.
With selective archive extraction, the link-editor makes multiple passes through an
archive to extract relocatable objects as needed to satisfy the symbol information being
accumulated in the link-editor internal symbol table. After the link-editor has made a
complete pass through the archive without extracting any relocatable objects, it moves
on to process the next input file.
Chapter 2 Link-Editor 29
By extracting from the archive only the relocatable objects needed at the time the
archive was encountered, the position of the archive within the input file list can be
significant. See Position of an Archive on the Command Line on page 32.
Note Although the link-editor makes multiple passes through an archive to resolve
symbols, this mechanism can be quite costly for large archives containing random
organizations of relocatable objects. In these cases, you should use tools like
lorder(1) and tsort(1) to order the relocatable objects within the archive and so
reduce the number of passes the link-editor must carry out.
The shared objects program data sections and most of the link-editing information
sections are unused by the link-editor. These sections are interpreted by the runtime
linker when the shared object is bound to generate a runnable process. However, the
occurrence of a shared object is remembered, and information is stored in the output
file image to indicate that this object is a dependency and must be made available at
runtime.
If a shared object has dependencies on other shared objects, these dependencies are
also processed. This processing occurs after all command-line input files have been
processed. These shared objects will be used to complete the symbol resolution
process; however, their names will not be recorded as dependencies in the output file
image being generated.
Although the position of a shared object on the link-edit command-line has less
significance than it does for archive processing, the position can have a global effect.
Multiple symbols of the same name are allowed to occur between relocatable objects
and shared objects, and between multiple shared objects. See Symbol Resolution
on page 38.
Note Multiple symbol definitions, and thus the information to describe the
interposing of one definition of a symbol for another, are reported in the load map
output generated using the -m option.
These conventions are recognized by the -l option of the link-editor. This option is
commonly used to supply additional libraries to a link-edit. The following example
directs the link-editor to search for libfoo.so. If the link-editor does not find
libfoo.so, it searches for libfoo.a before moving on to the next directory to be
searched.
$ cc -o prog file1.c file2.c -lfoo
Note There is a naming convention regarding the compilation environment and the
runtime environment use of shared objects. The compilation environment uses the
simple .so suffix, whereas the runtime environment commonly uses the suffix with an
additional version number. See Naming Conventions on page 102 and
Coordination of Versioned Filenames on page 146.
When link-editing in dynamic mode, you can choose to link with a mix of shared
objects and archives. When link-editing in static mode, only archive libraries are
acceptable for input.
Chapter 2 Link-Editor 31
When in dynamic mode and using the -l option to enable a library search, the
link-editor will first search in a given directory for a shared object that matches the
specified name. If no match is found, the link-editor looks for an archive library in the
same directory. When in static mode and using the -l option, only archive libraries
are sought.
By specifying the -B dynamic and -B static options on the command line as many
times as required, you can toggle the library search between shared objects or archives
respectively. For example, to link an application with the archive libfoo.a and the
shared object libbar.so, issue the following command:
$ cc -o prog main.o file1.c -Bstatic -lfoo -Bdynamic -lbar
The -B static and -B dynamic keywords are not exactly symmetrical. When you
specify -B static, the link-editor does not accept shared objects as input until the
next occurrence of -B dynamic. However, when you specify -B dynamic, the
link-editor first looks for shared objects and then archive librarys in any given
directory.
The precise description of the previous example is that the link-editor first searches for
libfoo.a, and then for libbar.so, and if that search fails, for libbar.a. Finally, it
searches for libc.so, and if that search fails, libc.a.
Therefore by default, the archive is not available to resolve any new references from
the input files that follow the archive on the command line. For example, the following
command directs the link-editor to search libfoo.a only to resolve symbol
references that have been obtained from file1.c. The libfoo.a archive is not
available to resolve symbol references from file2.c or file3.c.
$ cc -o prog file1.c -Bstatic -lfoo file2.c file3.c -Bdynamic
In some instances users have interdependencies between archives such that the
extraction of members from one archive is resolved by extracting members from
another archive. If these dependencies are cyclic, the archives must be specified
repeatedly on the command line to satisfy previous references. For example:
$ cc -o prog .... -lA -lB -lC -lA -lB -lC -lA
You can change the link-editor search path in two ways: using a command-line option,
or using an environment variable.
Path names defined using the -L option are used only by the link-editor. These path
names are not recorded in the output file image created for use by the runtime linker.
Chapter 2 Link-Editor 33
Note You must specify -L if you want the link-editor to search for libraries in your
current directory. You can use a period (.) to represent the current directory.
You can use the -Y option to change the default directories searched by the link-editor.
The argument supplied with this option takes the form of a colon separated list of
directories. For example, the following command searches for libfoo only in the
directories /opt/COMPILER/lib and /home/me/lib.
$ cc -o prog main.c -YP,/opt/COMPILER/lib:/home/me/lib -lfoo
The directories specified using the -Y option can be supplemented by using the -L
option.
The following example shows the combined effect of setting LD_LIBRARY_PATH and
calling the link-editor with several -L occurrences:
$ LD_LIBRARY_PATH=dir1:dir2;dir3
$ export LD_LIBRARY_PATH
$ cc -o prog main.c -Lpath1 ... -Lpath2 ... -Lpathn -lfoo
Note This environment variable can also be used to augment the search path of the
runtime linker. See Directories Searched by the Runtime Linker on page 64. To
prevent this environment variable from influencing the link-editor, use the -i option.
When a dynamic executable or shared object is linked with additional shared objects,
these shared objects are recorded as dependencies. These dependencies must be
located during process execution by the runtime linker. During the link-edit, one or
more search paths can be recorded in the output file. These search paths are used by
the runtime linker to locate any dependencies. These recorded search paths are
referred to as a runpath.
Specialized objects may be built with the -z nodefaultlib option to suppress any
search of the default location at runtime. Use of this option implies that all the
dependencies of an object can be located using its runpaths. Without this option, no
matter how you augment the runtime linkers search path, its last element is always
the default location. /usr/lib for 32bit objects and /usr/lib/64 for 64bit objects.
Note The default search path can be administrated using a runtime configuration file.
See Configuring the Default Search Paths on page 67. However, the creator of an
object should not rely on the existence of this file. You should always ensure that an
object can locate its dependencies with only its runpaths or the default location.
You can use the -R option, which takes a colon-separated list of directories, to record a
runpath in a dynamic executable or shared object. The following example records the
runpath /home/me/lib:/home/you/lib in the dynamic executable prog.
$ cc -o prog main.c -R/home/me/lib:/home/you/lib -Lpath1 \
-Lpath2 file1.c file2.c -lfoo -lbar
The runtime linker uses these paths, followed by the default location, to obtain any
shared object dependencies. In this case, this runpath is used to locate libfoo.so.1
and libbar.so.1.
For objects that may be installed in various locations, the $ORIGIN dynamic string
token provides a flexible means of recording a runpath. See Locating Associated
Dependencies on page 323.
Chapter 2 Link-Editor 35
Note A historic alternative to specifying the -R option is to set the environment
variable LD_RUN_PATH, and make this available to the link-editor. The scope and
function of LD_RUN_PATH and -R are identical, but when both are specified, -R
supersedes LD_RUN_PATH.
The sections .init and .fini provide a runtime initialization and termination code
block, respectively. However, the compiler drivers typically supply .init and .fini
sections with files they add to the beginning and end of your input file list. These files
have the effect of encapsulating the .init and .fini code into individual functions.
These functions are identified by the reserved symbol names _init and _fini
respectively. When creating a dynamic object, the link-editor identifies these symbols
with the .dynamic tags DT_INIT and DT_FINI accordingly. These tags identify the
associated sections so they may be called by the runtime linker.
For more information regarding the execution of initialization and termination code at
runtime see Initialization and Termination Routines on page 76.
The registration of initialization and termination functions can be carried out directly
by the link-editor using the -z initarray and -z finiarray options. For
example, the following command places the address of foo() in an .initarray
element, and the address of bar() in a .finiarray element.
$ cat main.c
#include <stdio.h>
void foo()
{
(void) printf("initializing: foo()\n");
}
void bar()
main()
{
(void) printf("main()\n");
return (0);
}
The creation of initialization and termination sections can be carried out directly using
an assembler. However, most compilers offer special primitives to simplify their
declaration. For example, the previous code example can be rewritten using the
following #pragma definitions. These definitions result in a call to foo() being
placed in an .init section, and a call to bar() being placed in a .fini section.
$ cat main.c
#include <stdio.h>
.......
$ cc -o main main.c
$ main
initializing: foo()
main()
finalizing: bar()
Initialization and termination code, spread throughout several relocatable objects, can
result in different behavior when included in an archive library or shared object. The
link-edit of an application using this archive might extract only a fraction of the objects
contained in the archive. These objects might provide only a portion of the
initialization and termination code spread throughout the members of the archive. At
runtime, only this portion of code is executed. The same application built against the
shared object will have all the accumulated initialization and termination code
executed when the dependency is loaded at runtime.
Chapter 2 Link-Editor 37
Symbol Processing
During input file processing, all local symbols from the input relocatable objects are
passed through to the output file image. All global symbols are accumulated internally
within the link-editor. Each global symbol supplied by a relocatable object is searched
for within this internal symbol table. If a symbol with the same name has already been
encountered from a previous input file, a symbol resolution process is called. This
symbol resolution process determines which of the two entries is kept.
On completing input file processing, and providing no fatal error conditions have
been encountered during symbol resolution, the link-editor determines if any
unresolved symbol references remain. Unresolved symbol references can cause the
link-edit to terminate.
Finally, the link-editors internal symbol table is added to the symbol tables of the
image being created.
The following sections expand upon symbol resolution and undefined symbol
processing.
Symbol Resolution
Symbol resolution runs the entire spectrum, from simple and intuitive to complex and
perplexing. Resolutions can be carried out silently by the link-editor, can be
accompanied by warning diagnostics, or can result in a fatal error condition.
The resolution of two symbols depends on their attributes, the type of file providing
the symbol, and the type of file being generated. For a complete description of symbol
attributes, see Symbol Table Section on page 222. For the following discussions,
however, it is worth identifying three basic symbol types:
Undefined Symbols that have been referenced in a file but have not been assigned
a storage address.
Tentative Symbols that have been created within a file but have not yet been sized
or allocated in storage. They appear as uninitialized C symbols, or FORTRAN
COMMON blocks within the file.
Defined Symbols that have been created and assigned storage addresses and space
within the file.
In its simplest form, symbol resolution involves the use of a precedence relationship
that has defined symbols dominating tentative symbols, which in turn dominate
undefined symbols.
int t_bar;
int d_bar = 1;
d_foo()
{
return (u_foo(u_bar, t_bar, d_bar));
}
$ cc -o main.o -c main.c
$ nm -x main.o
Simple Resolutions
Simple symbol resolutions are by far the most common, and result when two symbols
with similar characteristics are detected and one symbol takes precedence over the
other. This symbol resolution is carried out silently by the link-editor. For example, for
symbols with the same binding, a reference to an undefined symbol from one file is
bound to, or satisfied by, a defined or tentative symbol definition from another file. Or,
a tentative symbol definition from one file is bound to a defined symbol definition
from another file.
Symbols that undergo resolution can have either a global or weak binding. Weak
bindings have lower precedence than global binding, so symbols with different
bindings are resolved according to a slight alteration of the basic rules.
Weak symbols can usually be defined via the compiler, either individually or as aliases
to global symbols. One mechanism uses a #pragma definition:
$ cat main.c
#pragma weak bar
#pragma weak foo = _foo
int bar = 1;
_foo()
{
return (bar);
Chapter 2 Link-Editor 39
}
$ cc -o main.o -c main.c
$ nm -x main.o
[Index] Value Size Type Bind Other Shndx Name
...............
[7] |0x00000000|0x00000004|OBJT |WEAK |0x0 |3 |bar
[8] |0x00000000|0x00000028|FUNC |WEAK |0x0 |2 |foo
[9] |0x00000000|0x00000028|FUNC |GLOB |0x0 |2 |_foo
Notice that the weak alias foo is assigned the same attributes as the global symbol
_foo. This relationship is maintained by the link-editor and results in the symbols
being assigned the same value in the output image. In symbol resolution, weak
defined symbols are silently overridden by any global definition of the same name.
The function fread(3C), for example, is an ANSI C library function, whereas the
system function read(2) is not. A conforming ANSI C program must be able to
redefine read(2) and still use fread(3C) in a predictable way.
The problem here is that read(2) underlies the fread(3C) implementation in the
standard C library. Therefore, a program that redefines read(2) might confuse the
fread(3C) implementation. To guard against this occurrence, ANSI C states that an
implementation cannot use a name that is not reserved for it. Using the following
#pragma directive you can define just such a reserved name, and from it generate an
alias for the function read(2).
#pragma weak read = _read
Thus, you can quite freely define your own read() function without compromising
the fread(3C) implementation, which in turn is implemented to use the _read()
function.
The link-editor will not have difficulty with your redefinition of read(), either when
linking against the shared object or archive version of the standard C library. In the
former case, interposition takes its course. In the latter case, the fact that the C librarys
definition of read(2) is weak allows that definition to be quietly overridden.
You can use the link-editors -m option to write a list of all interposed symbol
references, along with section load address information, to the standard output.
$ cat bar.c
int array[2] = { 1, 2 };
Another form of attribute difference is the symbols type. In the following example the
symbol bar() has been defined as both a data item and a function.
$ cat foo.c
bar()
{
return (0);
}
$ cc -o libfoo.so -G -K pic foo.c
$ cat main.c
int bar = 1;
main()
{
return (bar);
}
$ cc -o main main.c -L. -lfoo
ld: warning: symbol bar has differing types:
(file main.o type=OBJT; file ./libfoo.so type=FUNC);
main.o definition taken
Note Symbol types in this context are classifications that can be expressed in ELF.
They are not related to the data types as employed by the programming language,
except in the crudest fashion.
Chapter 2 Link-Editor 41
In cases like the previous example, the relocatable object definition is taken when the
resolution occurs between a relocatable object and a shared object, or the first
definition is taken when the resolution occurs between two shared objects. When such
resolutions occur between symbols of different bindings (weak or global), a warning is
also produced.
Fatal Resolutions
Symbol conflicts that cannot be resolved result in a fatal error condition. In this case,
an appropriate error message is provided indicating the symbol name together with
the names of the files that provided the symbols, and no output file is generated.
Although the fatal condition is sufficient to terminate the link-edit, all input file
processing is first completed. In this manner, all fatal resolution errors can be
identified.
The most common fatal error condition exists when two relocatable objects both define
symbols of the same name, and neither symbol is a weak definition:
$ cat foo.c
int bar = 1;
$ cat bar.c
bar()
{
return (0);
}
foo.c and bar.c have conflicting definitions for the symbol bar. Because the
link-editor cannot determine which should dominate, the link-edit usually terminates
with an error message. You can use the link-editors -z muldefs option to suppress
this error condition, and allow the first symbol definition to be taken.
Undefined Symbols
After all of the input files have been read and all symbol resolution is complete, the
link-editor searches the internal symbol table for any symbol references that have not
been bound to symbol definitions. These symbol references are referred to as undefined
symbols. The effect of these undefined symbols on the link-edit process can vary
according to the type of output file being generated, and possibly the type of symbol.
main()
{
return (foo());
}
$ cc -o prog main.c
Undefined first referenced
symbol in file
foo main.o
ld: fatal: Symbol referencing errors. No output written to prog
In a similar manner, a symbol reference within a shared object that is never matched to
a symbol definition when the shared object is being used to create a dynamic
executable will also result in an undefined symbol:
$ cat foo.c
extern int bar;
foo()
{
return (bar);
}
If you want to allow undefined symbols, as in cases like the previous example, then
the default fatal error condition can be suppressed by using the link-editors
-z nodefs option.
Note Take care when using the -z nodefs option. If an unavailable symbol
reference is required during the execution of a process, a fatal runtime relocation error
occurs. It may be possible to detect this error during the initial execution and testing of
an application. However, more complex execution paths can result in this error
condition taking much longer to detect, which can be time consuming and costly.
Chapter 2 Link-Editor 43
Symbols can also remain undefined when a symbol reference in a relocatable object is
bound to a symbol definition in an implicitly defined shared object. For example,
continuing with the files main.c and foo.c used in the previous example:
$ cat bar.c
int bar = 1;
For this reason, bindings of this type are deemed fatal, and the implicit reference must
be made explicit by referencing the library directly during the link-edit of prog. The
required reference is hinted at in the fatal error message shown in the preceding
example.
The link-editors -z defs option can be used to force a fatal error if any undefined
symbols remain. This option is recommended when creating any shared objects.
Shared objects that reference symbols from an application can use the -z defs option
and define the applications symbols using the extern mapfile directive, as
described in Defining Additional Symbols on page 46.
A self-contained shared object, in which all references to external symbols are satisfied
by named dependencies, provides maximum flexibility. The shared object can be
employed by many users without those users having to determine and establish
dependencies to satisfy the shared objects requirements.
if ((fptr = foo) != 0)
(* fptr)(path);
}
When an application is built that references libfoo.so.1, the link-edit will complete
successfully regardless of whether a definition for the symbol foo is found. If during
execution of the application the function address tests nonzero, the function is called.
However, if the symbol definition is not found, the function address tests zero and so
it is not called.
Note Undefined weak references in this manner are discouraged. Instead, you should
use dlsym(3DL) with the RTLD_DEFAULT flag as a means of testing for a symbols
existence. See Testing for Functionality on page 91.
Chapter 2 Link-Editor 45
Tentative Symbol Order Within the Output File
Contributions from input files usually appear in the output file in the order of their
contribution. An exception occurs when processing tentative symbols and their
associated storage. These symbols are not fully defined until their resolution is
complete. If the resolution occurs as a result of encountering a defined symbol from a
relocatable object, then the order of appearance is that which would have occurred for
the definition.
If you need to control the ordering of a group of symbols, then any tentative definition
should be redefined to a zero-initialized data item. For example, the following
tentative definitions result in a reordering of the data items within the output file,
compared to the original order described in the source file foo.c:
$ cat foo.c
char A_array[0x10];
char B_array[0x20];
char C_array[0x30];
By defining these symbols as initialized data items, the relative ordering of these
symbols within the input file is carried over to the output file:
$ cat foo.c
char A_array[0x10] = { 0 };
char B_array[0x20] = { 0 };
char C_array[0x30] = { 0 };
For example, perhaps you want to generate a dynamic executable from the relocatable
object main.o, which refers to the symbols foo and bar. You want to obtain the
symbol definition foo from the relocatable object foo.o contained in lib1.a, and
the symbol definition bar from the relocatable object bar.o, contained in lib2.a.
However, the archive lib1.a also contains a relocatable object defining the symbol
bar. This relocatable object is presumably of differing functionality to the relocatable
object provided in lib2.a. To specify the required archive extraction, you can use the
following link-edit:
$ cc -o prog -L. -u foo -l1 main.o -l2
The -u option generates a reference to the symbol foo. This reference causes
extraction of the relocatable object foo.o from the archive lib1.a. The first reference
to the symbol bar occurs in main.o, which is encountered after lib1.a has been
processed. Therefore, the relocatable object bar.o is obtained from the archive
lib2.a.
Note This simple example assumes that the relocatable object foo.o from lib1.a
does not directly or indirectly reference the symbol bar. If it does then the relocatable
object bar.o is also extracted from lib1.a during its processing. See Archive
Processing on page 29 for a discussion of the link-editors multi-pass processing of an
archive.
A more extensive set of symbol definitions can be provided using the link-editors -M
option and an associated mapfile. The syntax for these mapfile entries is:
[ name ] {
scope:
symbol [ = [ type ] [ value ] [ size ] [ attribute ] ];
} [ dependency ];
name
A label for this set of symbol definitions, if present, identifies a version definition
within the image. See Chapter 5.
scope
Indicates the visibility of the symbols binding within the output file being
generated. All symbols defined with a mapfile are treated as global in scope
during the link-edit process. That is, they are resolved against any other symbols of
the same name obtained from any of the input files. The following definitions, and
aliases, define a symbols visibility in the object being created:
Chapter 2 Link-Editor 47
default / global
Symbols of this scope remain visible to other external objects. References to such
symbols from within the object are bound at runtime, thus allowing interposition
to take place.
protected / symbolic
Symbols of this scope remain visible to other external objects. References to these
symbols from within the object are bound at link-edit, thus preventing runtime
interposition. This scope definition has the same affect as a symbol with
STV_PROTECTED visibility. See Table 724.
hidden / local
Symbols of this scope are reduced to symbols with a local binding. Symbols of
this scope are not visible to other external objects. This scope definition has the
same affect as a symbol with STV_HIDDEN visibility. See Table 724.
eliminate
Symbols of this scope are hidden. Their symbol table entries are eliminated.
symbol
The name of the symbol required. If the name is not followed by one of the symbol
attributes, type, value, size or extern, a symbol reference is created. This reference is
exactly the same as would be generated using the -u option discussed earlier in
this section. If the symbol name is followed by any symbol attributes, then a
symbol definition is generated using the associated attributes.
When in local scope, this symbol name can be defined as the special auto-reduction
directive *. This directive results in all global symbols, not explicitly defined to be
global in the mapfile, receiving a local binding within any dynamic object file
being generated.
type
Indicates the symbol type attribute. This attribute can be either data, function,
or COMMON. The former two type attributes result in an absolute symbol definition.
See Symbol Table Section on page 222. The latter type attribute results in a
tentative symbol definition.
value
Indicates the value attribute and takes the form of Vnumber.
size
Indicates the size attribute and takes the form of Snumber.
attribute
This keyword provides additional attributes for the symbol:
EXTERN
Indicates the symbol is defined externally to the object being created. This
attribute can be used with the DIRECT or NODIRECT attributes to establish
individual direct, or no-direct references. Undefined symbols that would be
flagged with the -z defs option can also be suppressed with this option.
If the image being created is a relocatable object, then by default, no symbol reduction
is applied. In this case, any symbol reductions are recorded as part of the versioning
information. These reductions are applied when the relocatable object is finally used to
generate an executable or shared object. The link-editors -B reduce option can be
used to force symbol reduction when generating a relocatable object.
The following example shows how three symbol references can be defined. These
references are then used to extract members of an archive. Although this archive
extraction can be achieved by specifying multiple -u options to the link-edit, this
example also shows how the eventual scope of a symbol can be reduced to local.
$ cat foo.c
foo()
{
(void) printf("foo: called from lib.a\n");
}
$ cat bar.c
bar()
{
(void) printf("bar: called from lib.a\n");
}
$ cat main.c
extern void foo(), bar();
main()
Chapter 2 Link-Editor 49
{
foo();
bar();
}
$ ar -rc lib.a foo.o bar.o main.o
$ cat mapfile
{
local:
foo;
bar;
global:
main;
};
$ cc -o prog -M mapfile lib.a
$ prog
foo: called from lib.a
bar: called from lib.a
$ nm -x prog | egrep "main$|foo$|bar$"
[28] |0x00010604|0x00000024|FUNC |LOCL |0x0 |7 |foo
[30] |0x00010628|0x00000024|FUNC |LOCL |0x0 |7 |bar
[49] |0x0001064c|0x00000024|FUNC |GLOB |0x0 |7 |main
The significance of reducing symbol scope from global to local is covered in more
detail in the section Reducing Symbol Scope on page 52.
The following example shows how two absolute symbol definitions can be defined.
These definitions are then used to resolve the references from the input file main.c.
$ cat main.c
extern int foo();
extern int bar;
main()
{
(void) printf("&foo = %x\n", &foo);
(void) printf("&bar = %x\n", &bar);
}
$ cat mapfile
{
global:
foo = FUNCTION V0x400;
bar = DATA V0x800;
};
$ cc -o prog -M mapfile main.c
$ prog
&foo = 400 &bar = 800
$ nm -x prog | egrep "foo$|bar$"
[37] |0x00000800|0x00000000|OBJT |GLOB |0x0 |ABS |bar
[42] |0x00000400|0x00000000|FUNC |GLOB |0x0 |ABS |foo
When obtained from an input file, symbol definitions for functions or data items are
usually associated with elements of data storage. A mapfile definition is insufficient
to be able to construct this data storage, so these symbols must remain as absolute
values.
A feature of tentative symbols that differs from other symbol types is that their value
attribute indicates their alignment requirement. A mapfile definition can therefore be
used to realign tentative definitions obtained from the input files of a link-edit.
The following example shows the definition of two tentative symbols. The symbol foo
defines a new storage region whereas the symbol bar is actually used to change the
alignment of the same tentative definition within the file main.c.
$ cat main.c
extern int foo;
int bar[0x10];
main()
{
(void) printf("&foo = %x\n", &foo);
(void) printf("&bar = %x\n", &bar);
}
$ cat mapfile
{
global:
foo = COMMON V0x4 S0x200;
bar = COMMON V0x100 S0x40;
};
$ cc -o prog -M mapfile main.c
ld: warning: symbol bar has differing alignments:
(file mapfile value=0x100; file main.o value=0x4);
largest value applied
$ prog
&foo = 20940
&bar = 20900
$ nm -x prog | egrep "foo$|bar$"
[37] |0x00020900|0x00000040|OBJT |GLOB |0x0 |16 |bar
[42] |0x00020940|0x00000200|OBJT |GLOB |0x0 |16 |foo
Note This symbol resolution diagnostic can be suppressed by using the link-editors
-t option.
Chapter 2 Link-Editor 51
Reducing Symbol Scope
Symbol definitions defined to have local scope within a mapfile can be used to
reduce the symbols eventual binding. This mechanism can play an important role in
reducing the symbols visibility to future link-edits that use the generated file as part
of their input. In fact, this mechanism can provide for the precise definition of a files
interface, and so restrict the functionality made available to others.
For example, say you want to generate a simple shared object from the files foo.c
and bar.c. The file foo.c contains the global symbol foo, which provides the
service that you want to make available to others. The file bar.c contains the symbols
bar and str, which provide the underlying implementation of the shared object. The
creation of a simple shared object usually results in all three of these symbols having
global scope.
$ cat foo.c
extern const char * bar();
You can now use the functionality offered by this shared object as part of the link-edit
of another application. References to the symbol foo are bound to the implementation
provided by the shared object.
Because of their global binding, direct reference to the symbols bar and str is also
possible. This can have dangerous consequences, as you might later change the
implementation underlying the function foo. In so doing, you could unintentionally
cause an existing application that had bound to bar or str to fail or misbehave.
Another consequence of the global binding of the symbols bar and str is that they
can be interposed upon by symbols of the same name. The interposition of symbols
within shared objects is covered in section Simple Resolutions on page 39. This
interposition can be intentional and be used as a means of circumventing the intended
functionality offered by the shared object. On the other hand, this interposition can be
unintentional, the result of the same common symbol name used for both the
application and the shared object.
This symbol scope reduction has an additional performance advantage. The symbolic
relocations against the symbols bar and str that would have been necessary at
runtime are now reduced to relative relocations. This reduces the runtime overhead of
initializing and processing the shared object. See When Relocations are Performed
on page 121 for details of symbolic relocation overhead.
As the number of symbols processed during a link-edit increases, the ability to define
each local scope reduction within a mapfile becomes harder to maintain. An
alternative and more flexible mechanism enables you to define the shared objects
interface in terms of the global symbols that should be maintained, and instructs the
link-editor to reduce all other symbols to local binding. This mechanism is achieved
using the special auto-reduction directive *. For example, the previous mapfile
definition can be rewritten to define foo as the only global symbol required in the
output file generated:
$ cat mapfile
lib.so.1.1
{
global:
foo;
local:
*;
};
$ cc -o lib.so.1 -M mapfile -G foo.c bar.c
$ nm -x lib.so.1 | egrep "foo$|bar$|str$"
[30] |0x00000370|0x00000028|FUNC |LOCL |0x0 |6 |bar
[31] |0x00010428|0x00000004|OBJT |LOCL |0x0 |12 |str
[35] |0x00000348|0x00000028|FUNC |GLOB |0x0 |6 |foo
Chapter 2 Link-Editor 53
This example also defines a version name, lib.so.1.1, as part of the mapfile
directive. This version name establishes an internal version definition that defines the
files symbolic interface. The creation of a version definition is recommended. The
definition forms the foundation of an internal versioning mechanism that can be used
throughout the evolution of the file. See Chapter 5.
Note If a version name is not supplied, the output file name is used to label the
version definition. The versioning information created within the output file can be
suppressed using the link-editors -z noversion option.
The -B local option can be used to assert the auto-reduction directive * from the
command line. The previous example could be compiled successfully with:
$ cc -o lib.so.1 -M mapfile -B local -G foo.c bar.c
When generating an executable or shared object, any symbol reduction results in the
recording of version definitions within the output image, together with the reduction
of the appropriate symbols. When generating a relocatable object, the version
definitions are created but the symbol reductions are not processed. The result is that
the symbol entries for any symbol reductions still remain global. For example, using
the previous mapfile with the auto-reduction directive and associated relocatable
objects, an intermediate relocatable object is created that shows no symbol reduction.
$ cat mapfile
lib.so.1.1 {
global:
foo;
local:
*;
};
$ ld -o lib.o -M mapfile -r foo.o bar.o
$ nm -x lib.o | egrep "foo$|bar$|str$"
[17] |0x00000000|0x00000004|OBJT |GLOB |0x0 |3 |str
[19] |0x00000028|0x00000028|FUNC |GLOB |0x0 |1 |bar
The version definitions created within this image show that symbol reductions are
required. When the relocatable object is used eventually to generate an executable or
shared object, the symbol reductions occur. In other words, the link-editor reads and
interprets symbol reduction information contained in relocatable objects in the same
manner as it processes the data from a mapfile.
Thus, the intermediate relocatable object produced in the previous example can now
be used to generate a shared object:
$ ld -o lib.so.1 -G lib.o
$ nm -x lib.so.1 | egrep "foo$|bar$|str$"
[22] |0x000104a4|0x00000004|OBJT |LOCL |0x0 |14 |str
[24] |0x000003dc|0x00000028|FUNC |LOCL |0x0 |8 |bar
[36] |0x000003b4|0x00000028|FUNC |GLOB |0x0 |8 |foo
Symbol Elimination
An extension to symbol reduction is the elimination of a symbol entry from an objects
symbol table. Local symbols are only maintained in an objects .symtab symbol table.
This entire table can be removed from the object using the link-editors -s option, or
strip(1). On occasion, you might want to maintain the .symtab symbol table but
remove selected local symbol definitions from it.
Symbol elimination can be carried out using the mapfile directive eliminate. As
with the local directive, symbols can be individually defined, or the symbol name
can be defined as the special auto-elimination directive *. The following example
shows the elimination of the symbol bar for the previous symbol reduction example.
$ cat mapfile
lib.so.1.1
{
global:
foo;
local:
str;
eliminate:
*;
};
$ cc -o lib.so.1 -M mapfile -G foo.c bar.c
Chapter 2 Link-Editor 55
$ nm -x lib.so.1 | egrep "foo$|bar$|str$"
[31] |0x00010428|0x00000004|OBJT |LOCL |0x0 |12 |str
[35] |0x00000348|0x00000028|FUNC |GLOB |0x0 |6 |foo
The -B eliminate option can be used to assert the auto-elimination directive * from
the command line.
External Bindings
When a symbol reference from the object being created is satisfied by a definition
within a shared object, the symbol remains undefined. The relocation information
associated with the symbol provides for its lookup at runtime. The shared object that
provided the definition typically becomes a dependency.
The runtime linker employs a default search model to locate this definition at runtime.
It typically searches each object, starting with the dynamic executable, and progressing
through each dependency in the same order in which the objects were loaded.
Objects can also be created to use direct bindings. With this technique, the relationship
between the symbol reference and the object that provides the symbol definition is
maintained within the object being created. The runtime linker uses this information
to directly bind the reference to the object that defines the symbol, thus bypassing the
default symbol search model. See Direct Binding on page 69.
Linking objects that contribute a very large number of symbols may increase the
link-edit time due to the string table compression. To avoid this cost during
development use the link-editors -z nocompstrtab option. Any string table
compression performed during a link-edit can be displayed using the link-editors
debugging tokens -D strtab,detail.
Also included are any output relocation and dynamic information sections required by
the runtime linker. After all the output section information has been established, the
total output file size is calculated and the output file image is created accordingly.
When creating a dynamic executable or shared object, two symbol tables are usually
generated. The .dynsym table and its associated string table .dynstr contain register
(even if these are local), global, weak, and section symbols. These sections become part
of the text segment that is mapped as part of the process image at runtime (see the
mmap(2) man page). This enables the runtime linker to read these sections and perform
any necessary relocations.
The .symtab table, and its associated string table .strtab contain all the symbols
collected from the input file processing. These sections are not mapped as part of the
process image. They can even be stripped from the image using the link-editors -s
option, or after the link-edit using strip(1).
During the generation of the symbol tables, reserved symbols are created. These
symbols have special meaning to the linking process and should not be defined in
your code.
_etext
The first location after the text segment.
_edata
The first location after initialized data.
_end
The first location after all data.
_DYNAMIC
The address of the dynamic information section (the .dynamic section).
_END_
The same as _end. The symbol has local scope and, together with _START_,
provides a means of establishing an objects address range.
_GLOBAL_OFFSET_TABLE_
The position-independent reference to a link-editor supplied table of addresses, the
.got section. This table is constructed from position-independent data references
Chapter 2 Link-Editor 57
occurring in objects that have been compiled with the -K pic option. See
Position-Independent Code on page 115.
_PROCEDURE_LINKAGE_TABLE_
The position-independent reference to a link-editor supplied table of addresses, the
.plt section. This table is constructed from position-independent function
references occurring in objects that have been compiled with the -K pic option.
See Position-Independent Code on page 115.
_START_
The first location within the text segment. The symbol has local scope and, together
with _END_, provides a means of establishing an objects address range.
When generating an executable, the link-editor looks for additional symbols to define
the executables entry point. If a symbol was specified using the link-editors -e
option, that symbol is used. Otherwise the link-editor looks for the reserved symbol
names _start, and then main. If none of these symbols exists, the first address of the
text segment is used.
Relocation Processing
After you have created the output file, all data sections from the input files are copied
to the new image. Any relocations specified by the input files are applied to the output
image. Any additional relocation information that must be generated is also written to
the new image.
Displacement Relocations
Error conditions might occur if displacement relocations are applied to a data item,
which itself can be used in a copy relocation. The details of copy relocations are
covered in Copy Relocations on page 122.
A displacement relocation remains valid when both the relocated offset and the target
to which it is relocated remain separated by the same displacement. A copy relocation
is one where a global data item within a shared object is copied to the .bss of an
To help diagnose these problem areas, the link-editor indicates the displacement
relocation use of a dynamic object with one or more dynamic DT_FLAGS_1 flags, as
shown in Table 744. In addition, the link-editors -z verbose option can be used to
display suspicious relocations.
For example, say you create a shared object with a global data item, bar[], which has
a displacement relocation applied to it. This item could be copy-relocated if referenced
from a dynamic executable. The link-editor warns of this condition with:
$ cc -G -o libfoo.so.1 -z verbose -K pic foo.o
ld: warning: relocation warning: R_SPARC_DISP32: file foo.o: symbol foo: \
displacement relocation to be applied to the symbol bar: at 0x194: \
displacement relocation will be visible in output image
If you now create an application that references the data item bar[], a copy relocation
will be created which results in the displacement relocation being invalidated. Because
the link-editor can explicitly discover this situation, an error message is generated
regardless of the use of the -z verbose option.
$ cc -o prog prog.o -L. -lfoo
ld: warning: relocation error: R_SPARC_DISP32: file foo.so: symbol foo: \
displacement relocation applied to the symbol bar at: 0x194: \
the symbol bar is a copy relocated symbol
Note ldd(1), when used with either the -d or -r options, uses the displacement
dynamic flags to generate similar relocation warnings.
Chapter 2 Link-Editor 59
These error conditions can be avoided by ensuring that the symbol definition being
relocated (offset) and the symbol target of the relocation are both local. Use static
definitions or the link-editors scoping technology. See Reducing Symbol Scope
on page 52. Relocation problems such as these can be avoided by accessing data
within shared objects using functional interfaces.
Debugging Aids
A debugging library is provided with the Solaris linkers. This library enables you to
trace the link-editing process in more detail. This library helps you understand, or
debug, the link-edit of your own applications or libraries. Although the type of
information displayed using this library is expected to remain constant, the exact
format of the information might change slightly from release to release.
Some of the debugging output might be unfamiliar if you do not have an intimate
knowledge of the ELF format. However, many aspects might be of general interest to
you.
Debugging is enabled by using the -D option, and all output produced is directed to
the standard error. This option must be augmented with one or more tokens to
indicate the type of debugging required. The tokens available can be displayed by
typing -D help at the command line.
$ ld -Dhelp
debug:
debug: For debugging the link-editing of an application:
debug: LD_OPTIONS=-Dtoken1,token2 cc -o prog ...
debug: or,
debug: ld -Dtoken1,token2 -o prog ...
debug: where placement of -D on the command line is significant
debug: and options can be switched off by prepending with !.
debug:
debug:
debug: args display input argument processing
debug: basic provide basic trace information/warnings
debug: detail provide more information in conjunction with other options
debug: entry display entrance criteria descriptors
debug: files display input file processing (files and libraries)
debug: got display GOT symbol information
debug: help display this help message
debug: libs display library search paths; detail flag shows actual
debug: library lookup (-l) processing
debug: map display map file processing
debug: move display move section processing
debug: reloc display relocation processing
debug: sections display input section processing
debug: segments display available output segments and address/offset
Note This listing is an example, and shows the options meaningful to the link-editor.
The exact options might differ from release to release.
Most compiler drivers interpret the -D option during their preprocessing phase.
Therefore, the LD_OPTIONS environment variable is a suitable mechanism for passing
this option to the link-editor.
The following example shows how input files can be traced. This syntax can be
especially useful in determining what libraries have been located, or what relocatable
objects have been extracted from an archive during a link-edit.
$ LD_OPTIONS=-Dfiles cc -o prog main.o -L. -lfoo
............
debug: file=main.o [ ET_REL ]
debug: file=./libfoo.a [ archive ]
debug: file=./libfoo.a(foo.o) [ ET_REL ]
debug: file=./libfoo.a [ archive ] (again)
............
Here the member foo.o is extracted from the archive library libfoo.a to satisfy the
link-edit of prog. Notice that the archive is searched twice to verify that the extraction
of foo.o did not warrant the extraction of additional relocatable objects. More than
one (again) display indicates that the archive is a candidate for ordering using
lorder(1) and tsort(1).
By using the symbols token, you can determine which symbol caused an archive
member to be extracted, and which object made the initial symbol reference.
$ LD_OPTIONS=-Dsymbols cc -o prog main.o -L. -lfoo
............
debug: symbol table processing; input file=main.o [ ET_REL ]
............
debug: symbol[7]=foo (global); adding
debug:
debug: symbol table processing; input file=./libfoo.a [ archive ]
debug: archive[0]=bar
debug: archive[1]=foo (foo.o) resolves undefined or tentative symbol
debug:
debug: symbol table processing; input file=./libfoo(foo.o) [ ET_REL ]
Chapter 2 Link-Editor 61
.............
The symbol foo is referenced by main.o and is added to the link-editors internal
symbol table. This symbol reference causes the extraction of the relocatable object
foo.o from the archive libfoo.a.
By using the detail token together with the symbols token, the details of symbol
resolution during input file processing can be observed.
$ LD_OPTIONS=-Dsymbols,detail cc -o prog main.o -L. -lfoo
............
debug: symbol table processing; input file=main.o [ ET_REL ]
............
debug: symbol[7]=foo (global); adding
debug: entered 0x000000 0x000000 NOTY GLOB UNDEF REF_REL_NEED
debug:
debug: symbol table processing; input file=./libfoo.a [ archive ]
debug: archive[0]=bar
debug: archive[1]=foo (foo.o) resolves undefined or tentative symbol
debug:
debug: symbol table processing; input file=./libfoo.a(foo.o) [ ET_REL ]
debug: symbol[1]=foo.c
.............
debug: symbol[7]=bar (global); adding
debug: entered 0x000000 0x000004 OBJT GLOB 3 REF_REL_NEED
debug: symbol[8]=foo (global); resolving [7][0]
debug: old 0x000000 0x000000 NOTY GLOB UNDEF main.o
debug: new 0x000000 0x000024 FUNC GLOB 2 ./libfoo.a(foo.o)
debug: resolved 0x000000 0x000024 FUNC GLOB 2 REF_REL_NEED
............
The original undefined symbol foo from main.o has been overridden with the
symbol definition from the extracted archive member foo.o. The detailed symbol
information reflects the attributes of each symbol.
In the previous example, you can see that using some of the debugging tokens can
produce a wealth of output. In cases where you are interested only in the activity
around a subset of the input files, the -D option can be placed directly in the link-edit
command-line, and toggled on and off. In the following example the display of symbol
processing is switched on only during the processing of the library libbar.
$ ld .... -o prog main.o -L. -Dsymbols -lbar -D!symbols ....
Note To obtain the link-edit command line you might have to expand the
compilation line from any driver being used. See Using a Compiler Driver
on page 27.
Runtime Linker
During the process of executing a dynamic object the kernel loads the file and reads
the program header information. See Program Header on page 236. From this
information the kernel locates the name of the required interpreter. The kernel loads
this interpreter and transfers control to it, passing sufficient information to enable the
interpreter to continue binding the application and run it.
63
Calls any initialization functions provided by the dependencies.
Passes control to the application.
Can be called upon during the applications execution, to perform any delayed
function binding.
Can be called upon by the application to acquire additional objects with
dlopen(3DL), and bind to symbols within these objects with dlsym(3DL).
Note Even when a shared object is referenced multiple times in the dependency list,
the runtime linker connects the object only once to the process.
The runtime linker uses this dependency information to locate, and load, the
associated objects. These dependencies are processed in the same order as they were
referenced during the link-edit of the executable.
Once all the dynamic executables dependencies are loaded, they too are inspected, in
the order they are loaded, to locate any additional dependencies. This process
continues until all dependencies are located and loaded. This technique results in a
breadth-first ordering of all dependencies.
The file /usr/bin/cat has a dependency, or needs, the files libc.so.1 and
libdl.so.1.
The dependencies recorded in an object can be inspected using dump(1). Use this
command to display the files .dynamic section, and look for entries that have a
NEEDED tag. In the following example, the dependency libdl.so.1, displayed in the
previous ldd(1) example, is not recorded in the file /usr/bin/cat. ldd(1) shows the
total dependencies of the specified file, and libdl.so.1 is actually a dependency of
/usr/lib/libc.so.1.
$ dump -Lvp /usr/bin/cat
/usr/bin/cat:
[INDEX] Tag Value
[1] NEEDED libc.so.1
.........
In the previous dump(1) example, the dependencies are expressed as simple file names.
In other words, there is no / in the name. The use of a simple file name requires the
runtime linker to generate the path name from a set of rules. File names that contain
an embedded /, are used as provided.
The simple file name recording is the standard, most flexible mechanism of recording
dependencies. The -h option of the link-editor records a simple name within the
dependency. See Naming Conventions on page 102 and Recording a Shared Object
Name on page 103.
You can specify additional search path, on a per-object basis, by recording a runpath
during the link-edit of an object. See Directories Searched by the Runtime Linker
on page 35 for details on recording this information.
Any runpath recording can be displayed using dump(1). Reference the .dynamic
entry that has the RUNPATH tag. In the following example, prog has a dependency on
libfoo.so.1. The runtime linker must search directories /home/me/lib and
/home/you/lib before it looks in the default location.
$ dump -Lvp prog
prog:
[INDEX] Tag Value
[1] NEEDED libfoo.so.1
Another way to add to the runtime linkers search path is to set the environment
variable LD_LIBRARY_PATH. This environment variable, which is analyzed once at
process startup, can be set to a colon-separated list of directories. These directories are
searched by the runtime linker before any runpath specification or default directory.
These environment variables are well suited to debugging purposes, such as forcing
an application to bind to a local dependency. In the following example, the file prog
from the previous example is bound to libfoo.so.1, found in the present working
directory.
$ LD_LIBRARY_PATH=. prog
Environmental search paths can result in a 64bit executable searching a path that
contains a 32bit library matching the name being looked for. Or, vice versa. The
runtime linker rejects the mismatched 32bit library and continues down its search
path looking for a valid 64bit match. If no match is found, an error message is
generated. This can be observed in detail by setting the LD_DEBUG environment
variable to include the files token. See Debugging Library on page 93.
$ LD_LIBRARY_PATH=/usr/bin/64 LD_DEBUG=files /usr/bin/ls
...
00283: file=libc.so.1; needed by /usr/bin/ls
00283:
00283: file=/usr/lib/64/libc.so.1 rejected: ELF class mismatch: \
00283: 32-bit/64-bit
00283:
00283: file=/usr/lib/libc.so.1 [ ELF ]; generating link map
00283: dynamic: 0xef631180 base: 0xef580000 size: 0xb8000
00283: entry: 0xef5a1240 phdr: 0xef580034 phnum: 3
00283: lmid: 0x0
00283:
00283: file=/usr/lib/libc.so.1; analyzing [ RTLD_GLOBAL RTLD_LAZY ]
...
If a dependency cannot be located, ldd(1) indicates that the object cannot be found.
Any attempt to execute the application results in an appropriate error message from
the runtime linker:
$ ldd prog
libfoo.so.1 => (file not found)
libc.so.1 => /usr/lib/libc.so.1
Relocation Processing
After the runtime linker has loaded all the dependencies required by an application,
the linker processes each object and performs all necessary relocations.
During the link-editing of an object, any relocation information supplied with the
input relocatable objects is applied to the output file. However, when creating a
dynamic executable or shared object, many of the relocations cannot be completed at
For a more detailed description of the many relocation types, see Relocation Types
(Processor-Specific) on page 212. There are two basic types of relocations:
Non-symbolic relocations
Symbolic relocations
The relocation records for an object can be displayed by using dump(1). In the
following example, the file libbar.so.1 contains two relocation records that
indicate that the global offset table (the .got section) must be updated.
$ dump -rvp libbar.so.1
libbar.so.1:
.rela.got:
Offset Symndx Type Addend
0x10438 0 R_SPARC_RELATIVE 0
0x1043c foo R_SPARC_GLOB_DAT 0
The first relocation is a simple relative relocation that can be seen from its relocation
type and the symbol index (Symndx) field being zero. This relocation needs to use the
base address at which the object was loaded into memory to update the associated
.got offset.
The second relocation requires the address of the symbol foo. To complete this
relocation, the runtime linker must locate this symbol from either the dynamic
executable or one of its dependencies.
Symbol Lookup
When an object requires a symbol, the runtime linker searches for that symbol based
upon the requesting objects symbol search scope, and the symbol visibility offered by
each object within the process. These attributes are applied as defaults to an object at
the time the object is loaded, as specific modes to dlopen(3DL), and in some cases can
be recorded within the object at the time it is built.
Typically, an average user becomes familiar with the default symbol search models
that are applied to a dynamic executable and its dependencies, and to objects obtained
through dlopen(3DL). The former is outlined in the next section Default Lookup
on page 69, and the latter, which is also able to exploit the various symbol lookup
attributes, is discussed in Symbol Lookup on page 84.
Default Lookup
A dynamic executable and all the dependencies loaded with it are assigned world
search scope, and global symbol visibility. See Symbol Lookup on page 84. When
the runtime linker looks up a symbol for a dynamic executable or for any of the
dependencies loaded with the executable, it does so by searching each object. The
runtime linker starts with the dynamic executable, and progresses through each
dependency in the same order in which the objects were loaded.
The runtime linker first looks for foo in the dynamic executable prog, then in the
shared object /home/me/lib/libfoo.so.1, and finally in the shared object
/home/me/lib/libbar.so.1.
Note Symbol lookup can be an expensive operation, especially when the size of
symbol names increases and the number of dependencies increases. This aspect of
performance is discussed in more detail in Performance Considerations on page 112.
See Direct Binding on page 69 for an alternative lookup model.
Interposition
The runtime linkers default mechanism of searching for a symbol first in the dynamic
executable and then in each of the dependencies means that the first occurrence of the
required symbol will satisfy the search. Therefore, if more than one instance of the
same symbol exists, the first instance interposes on all others. See also Shared Object
Processing on page 30.
Direct Binding
When creating an object to use direct bindings, the relationship between the
referenced symbol and the dependency that provided the definition is recorded in the
object. The runtime linker uses this information to search directly for the symbol in the
The direct binding model can significantly reduce the symbol lookup overhead within
a dynamic process that has many symbolic relocations and many dependencies. This
model also enables multiple symbols of the same name to be located from different
objects that have been bound to directly.
Direct binding can circumvent the traditional use of interposition symbols because it
bypasses the default search model. The default model ensures that all references to a
symbol bind to one definition.
Note Direct bindings can be disabled at runtime by setting the environment variable
LD_NODIRECT to a non-null value.
Note NODIRECT mapfile directives can be combined with the command line
options -B direct or -z direct. Symbols that are not explicitly defined NODIRECT
will follow the command line directive.
Procedure linkage table entries are constructed so that when they are first called,
control is passed to the runtime linker. The runtime linker looks up the required
symbol and rewrites information in the associated object so that any future calls to this
procedure linkage table entry go directly to the function. This mechanism enables
relocations of this type to be deferred until the first instance of a function is called.
This process is sometimes referred to as lazy binding.
The runtime linkers default mode is to perform lazy binding whenever procedure
linkage table relocations are provided. This default can be overridden by setting the
environment variable LD_BIND_NOW to any non-null value. This environment variable
setting causes the runtime linker to perform both immediate and lazy reference
Objects can also be accessed with dlopen(3DL) with the mode defined as RTLD_NOW.
Objects can also be built using the link-editors -z now option to indicate that they
require complete relocation processing at the time they are loaded. This relocation
requirement is also propagated to any dependencies of the marked object at runtime.
Note Although the preceding examples of immediate and lazy references are typical,
the creation of procedure linkage table entries is ultimately controlled by the relocation
information provided by the relocatable object files used as input to a link-edit.
Relocation records such as R_SPARC_WPLT30 and R_386_PLT32 instruct the
link-editor to create a procedure linkage table entry are common for
position-independent code. However, as a dynamic executable has a fixed location,
external function references that can be determined at link-edit time can be converted
to procedure linkage table entries regardless of the original relocation records.
Relocation Errors
The most common relocation error occurs when a symbol cannot be found. This
condition results in an appropriate runtime linker error message and the termination
of the application. For example:
$ ldd prog
libfoo.so.1 => ./libfoo.so.1
libc.so.1 => /usr/lib/libc.so.1
libbar.so.1 => ./libbar.so.1
libdl.so.1 => /usr/lib/libdl.so.1
$ prog
ld.so.1: prog: fatal: relocation error: file ./libfoo.so.1: \
symbol bar: referenced symbol not found
The symbol bar, which is referenced in the file libfoo.so.1, cannot be located.
During the link-edit of a dynamic executable, any potential relocation errors of this
sort are flagged as fatal undefined symbols. See Generating an Executable Output
File on page 43 for examples. This runtime relocation error can occur if the link-edit
of main used a different version of the shared object libbar.so.1 that contained a
symbol definition for bar, or if the -z nodefs option was used as part of the
link-edit.
To guard against errors of this kind, the relocation requirements of any dynamic
executable or shared object can be validated using ldd(1).
When the -d option is specified with ldd(1), all dependencies will be printed and all
immediate reference relocations will be processed. If a reference cannot be resolved, a
diagnostic message is produced. From the previous example this option would result
in:
$ ldd -d prog
libfoo.so.1 => ./libfoo.so.1
libc.so.1 => /usr/lib/libc.so.1
libbar.so.1 => ./libbar.so.1
libdl.so.1 => /usr/lib/libdl.so.1
symbol not found: bar (./libfoo.so.1)
When the -r option is specified with ldd(1), all immediate and lazy reference
relocations are processed. If either type of relocation cannot be resolved, a diagnostic
message is produced.
The order in which these objects are processed can be displayed using ldd(1):
$ LD_PRELOAD=./newstuff.so.1 ldd prog
./newstuff.so.1 => ./newstuff.so
libc.so.1 => /usr/lib/libc.so.1
In another example the preloading is a little more complex and time consuming.
$ LD_PRELOAD="./foo.o ./bar.o" prog
These mechanisms of inserting an object after a dynamic executable take the concept
of interposition to another level. You can use these mechanisms to experiment with a
new implementation of a function that resides in a standard shared object. If you
preload an object containing this function, the object interposes on the original. Thus
the old functionality can be completely hidden with the new preloaded version.
Under this default model, all the dependencies of an application are loaded into
memory, and all data relocations are performed. These operations are performed
regardless of whether the code in these dependencies is referenced by the application
during its execution.
Under a lazy loading model, any dependencies that are labeled for lazy loading are
loaded only when explicitly referenced. By taking advantage of a function calls lazy
binding, the loading of a dependency is delayed until it is first referenced. In fact,
objects that are never referenced are never loaded.
The alternate method of achieving a lazy loading model is to use dlopen() and
dlsym() to load and bind to a dependency when needed. This is ideal if the number
of dlsym() references is small, or the dependency name or location is not known at
link-edit time. For more complex interactions with known dependencies, coding to
normal symbol references and designating the dependency to be lazily loaded is
simpler.
The POSFLAG_1 with the value of LAZY designates that the following NEEDED entry,
libdebug.so.1, should be lazily loaded. Because libc.so.1 has no preceding
LAZY flag it is loaded at the initial startup of the program.
The use of lazy loading can require a precise declaration of dependencies and
runpaths through out the objects used by an application. For example, suppose two
objects, libA.so and libB.so, both make reference to symbols in libX.so.
libA.so declares it has a dependency on libX.so, but libB.so does not. Typically,
when libA.so and libB.so are used together, libB.so can reference libX.so
because libA.so made it available. But, if libA.so declares libX.so to be lazy
Regardless of lazy loading, it is recommended that dynamic objects declare all their
dependencies and how to locate them. With lazy loading, this dependency
information becomes even more important.
Note Lazy loading can be disabled at runtime by setting the environment variable
LD_NOLAZYLOAD to a non-null value.
The runtime linker executes functions whose addresses are contained in the
.preinit_array and .init_array sections. These functions are executed in the
same order in which their addresses appear in the array. The runtime linker executes
an .init section as an individual function. If an object contains both .init and
.init_array sections, the .init section is processed before the functions defined
by the .init_array section for that object.
Note Any .init section within the dynamic executable is called from the
application itself by the process startup mechanism supplied by the compiler driver.
The .init section within the dynamic executable is called last, after all dependency
initialization sections are executed.
Dynamic objects can also provide termination sections. The termination sections
.fini_array and .fini are created by the link-editor when a dynamic object is
built.
The runtime linker executes functions whose addresses are contained in the
.fini_array section. These functions are executed in the reverse order in which
their addresses appear in the array. The runtime linker executes a .fini section as an
individual function. If an object contains both .fini and .fini_array sections, the
functions defined by the .fini_array section are processed before the .fini
section for that object.
Note Any .fini section within the dynamic executable is called from the
application itself by the process termination mechanism supplied by the compiler
driver. The .fini section of the dynamic executable is called first, before all
dependency termination sections are executed.
For more information regarding the creation of initialization and termination sections
by the link-editor see Initialization and Termination Sections on page 36.
Prior to the Solaris 2.6 release, dependency initialization routines were called in reverse
load order, which is the reverse order of the dependencies displayed with ldd(1).
Similarly, dependency termination routines were called in load order. However, as
dependency hierarchies became more complex, this simple ordering approach became
inadequate.
Starting with the Solaris 2.6 release, the runtime linker constructs a topologically
sorted list of objects that have been loaded. This list is built from the dependency
relationship expressed by each object, together with any symbol bindings that occur
outside of the expressed dependencies.
Use ldd(1) with the -i option to display the initialization order of an objects
dependencies. For example, the following dynamic executable and its dependencies
exhibit a cyclic dependency:
$ dump -Lv B.so.1 | grep NEEDED
[1] NEEDED C.so.1
$ dump -Lv C.so.1 | grep NEEDED
[1] NEEDED B.so.1
$ dump -Lv main | grep NEEDED
[1] NEEDED A.so.1
[2] NEEDED B.so.1
[3] NEEDED libc.so.1
$ ldd -i main
A.so.1 => ./A.so.1
B.so.1 => ./B.so.1
libc.so.1 => /usr/lib/libc.so.1
C.so.1 => ./C.so.1
libdl.so.1 => /usr/lib/libdl.so.1
init object=/usr/lib/libc.so.1
init object=./A.so.1
init object=./C.so.1 - cyclic group [1], referenced by:
./B.so.1
init object=./B.so.1 - cyclic group [1], referenced by:
./C.so.1
Caution Prior to Solaris 8 10/00, the environment variable LD_BREADTH could be set
to a non-null value to force the runtime linker to execute initialization and termination
sections in pre-Solaris 2.6 order. This functionality has since been disabled, as the
initialization dependencies of many applications have become complex and mandate
topological sorting. Any LD_BREADTH setting is now silently ignored.
Initialization processing is repeated for any objects added to the running process with
dlopen(3DL). Termination processing is also carried out for any objects unloaded
from the process as a result of a call to dlclose(3DL).
Dynamic initialization can not be revealed with ldd(1). However, the exact sequence
of initialization calls can be observed at runtime by setting the LD_DEBUG
environment variable to include the token basic. See Debugging Library on page 93.
Dynamic initialization is only available when processing lazy references. Use of the
environment variable LD_BIND_NOW, objects built with the -z now option, or objects
referenced by dlopen(3DL) with mode RTLD_NOW, circumvent any dynamic
initialization.
Note Objects that are pending initialization, and are referenced through
dlopen(3DL), will be initialized prior to returning control from this function.
Keep the content of initialization and termination sections to a minimum. Avoid global
constructors by initializing objects at runtime. Reduce the dependency of initialization
and termination code on other dependencies. Define the dependency requirements of
all dynamic objects. See Generating a Shared Object Output File on page 44. Do not
express dependencies that are not required. See Shared Object Processing
on page 30. Avoid cyclic dependencies. Do not depend on the order of an initialization
or termination sequence. The ordering of objects can be affected by both shared object
and application development. See Dependency Ordering on page 106.
The runtime linker categorizes a process as secure if the user is not a super-user, and
the real user and effective user identifiers are not equal. Similarly, if the user is not a
super-user and the real group and effective group identifiers are not equal, the process
is deemed secure. See the getuid(2), geteuid(2), getgid(2) and getegid(2) man
pages.
The default trusted directory known to the runtime linker is /usr/lib/secure for
32bit objects or /usr/lib/secure/64 for 64bit objects. The utility crle(1) may be
used to specify additional trusted directories applicable for secure applications.
Administrators who use this technique should ensure that the target directories are
suitably protected from malicious intrusion.
In a secure process, the expansion of the $ORIGIN string is allowed only if it expands
to a trusted directory. See Security on page 326.
Additional objects can be loaded with a secure process using the LD_PRELOAD or
LD_AUDIT environment variables. These objects must be specified as full path names
or simple file names. Full path names are restricted to known trusted directories.
Simple file names, in which no / appears in the name, are located subject to the
search path restrictions previously described. Simple file names resolve only to known
trusted directories.
In a secure process, any dependencies that consist of simple file names are processed
using the path name restrictions previously described. Dependencies expressed as full
or relative path names are used as is. Therefore, the developer of a secure process
should ensure that the target directory referenced as a full or relative path name
dependency is suitably protected from malicious intrusion.
An application can use the following typical scenario to access an additional shared
object.
A shared object is located and added to the address space of a running application
using dlopen(3DL). Any dependencies that this shared object has are located and
added at this time.
The added shared object and its dependencies are relocated. Any initialization
sections within these objects are called.
The application locates symbols within the added objects using dlsym(3DL). The
application can then reference the data or call the functions defined by these new
symbols.
After the application has finished with the objects, the address space can be freed
using dlclose(3DL). Any termination sections within the objects being freed is
called at this time.
Any error conditions that occur as a result of using these runtime linker interface
routines can be displayed using dlerror(3DL).
The services of the runtime linker are defined in the header file dlfcn.h and are
made available to an application by the shared object libdl.so.1. In the following
example, the file main.c can make reference to any of the dlopen(3DL) family of
routines, and the application prog can bind to these routines at runtime.
If the path name is specified as a simple file name, one with no / in the name, then
the runtime linker will use a set of rules to generate an appropriate path name. Path
names that contain a / will be used as provided.
These search path rules are exactly the same as are used to locate any initial
dependencies. See Directories Searched by the Runtime Linker on page 64. For
example, if the file main.c contains the following code fragment:
#include <stdio.h>
#include <dlfcn.h>
then to locate the shared object foo.so.1, the runtime linker uses any
LD_LIBRARY_PATH definition present at process initialization, followed by any
runpath specified during the link-edit of prog. Finally, the runtime linker uses the
default location /usr/lib for 32bit objects, and /usr/lib/64 for 64bit objects.
then the runtime linker searches for the file only in the current working directory of
the process.
Note Any shared object specified using dlopen(3DL) should be referenced by its
versioned file name. For more information on versioning, see Coordination of
Versioned Filenames on page 146.
If the object being added by dlopen(3DL) has dependencies on other objects, they too
are brought into the processs address space. This process continues until all the
dependencies of the specified object are loaded. This dependency tree is referred to as
a group.
If the object specified by dlopen(3DL), or any of its dependencies, are already part of
the process image, then the objects are not processed any further. A valid handle is
returned to the application. This mechanism prevents the same object from being
loaded more than once, and enables an application to obtain a handle to itself. For
example, if the previous main.c example contained the following dlopen() call:
if ((handle = dlopen((const char *)0, RTLD_LAZY)) == NULL) {
then the handle returned from dlopen(3DL) can be used to locate symbols within the
application itself, within any of the dependencies loaded as part of the processs
initialization, or within any objects added to the processs address space, using a
dlopen(3DL) that specified the RTLD_GLOBAL flag.
Relocation Processing
As described in Chapter 3, after locating and loading any objects, the runtime linker
must process each object and perform any necessary relocations. Any objects brought
into the processs address space with dlopen(3DL) must also be relocated in the same
manner.
For simple applications this process is straightforward. However, for users who have
more complex applications with many dlopen(3DL) calls involving many objects,
possibly with common dependencies, this process can be quite important.
Relocations can be categorized according to when they occur. The default behavior of
the runtime linker is to process all immediate reference relocations at initialization and
all lazy references during process execution, a mechanism commonly referred to as
lazy binding.
This same mechanism is applied to any objects added with dlopen(3DL) when the
mode is defined as RTLD_LAZY. An alternative is to require all relocations of an object
to be performed immediately when the object is added. You can use a mode of
RTLD_NOW, or record this requirement in the object when it is built using the
link-editors -z now option. This relocation requirement is propagated to any
dependencies of the object being opened.
Symbol Lookup
If an object acquired by dlopen(3DL) refers to a global symbol, the runtime linker
must locate this symbol from the pool of objects that make up the process. In the
absence of direct binding, a default symbol search model is applied to objects obtained
by dlopen(3DL). However, the mode of a dlopen(3DL), combined with the attributes
of the objects that make up the process, provide for alternative symbol search models.
Objects that required direct binding, although maintaining all the attributes described
later, search for symbols directly in the associated dependency. See Direct Binding
on page 69.
Two attributes of an object affect symbol lookup. The first is the requesting objects
symbol search scope, and the second is the symbol visibility offered by each object
within the process. An objects search scope can be:
world
The object can look in any other global object within the process.
group
The object can look only in an object of the same group. The dependency tree
created from an object obtained with dlopen(3DL), or from an object built using
the link-editors -B group option, forms a unique group.
By default, objects obtained with dlopen(3DL) are assigned world symbol search
scope, and local symbol visibility. The section, Default Symbol Lookup Model
on page 85, uses this default model to illustrate typical object group interactions. The
sections Defining a Global Object on page 88, Isolating a Group on page 88, and
Object Hierarchies on page 88 show examples of using dlopen(3DL) modes and
file attributes to extend the default symbol lookup model.
In the following example, the dynamic executable prog and the shared object B.so.1
each have the following (simplified) dependencies:
$ ldd prog
A.so.1 => ./A.so.1
$ ldd B.so.1
C.so.1 => ./C.so.1
If prog acquires the shared object B.so.1 by dlopen(3DL), then any symbol
required to relocate the shared objects B.so.1 and C.so.1 will first be looked for in
prog, followed by A.so.1, followed by B.so.1, and finally in C.so.1. In this
simple case, think of the shared objects acquired through the dlopen(3DL) as if they
had been added to the end of the original link-edit of the application. For example, the
objects referenced in the previous listing can be expressed diagrammatically as shown
in the following figure.
Any symbol lookup required by the objects acquired from the dlopen(3DL), shown as
shaded blocks, proceeds from the dynamic executable prog through to the final
shared object C.so.1.
This symbol lookup is established by the attributes assigned to the objects as they
were loaded. Recall that the dynamic executable and all the dependencies loaded with
it are assigned global symbol visibility, and that the new objects are assigned world
symbol search scope. Therefore, the new objects are able to look for symbols in the
original objects. The new objects also form a unique group in which each object has
local symbol visibility. Therefore, each object within the group can look for symbols
within the other group members.
These new objects do not affect the normal symbol lookup required by either the
application or its initial object dependencies. For example, if A.so.1 requires a
function relocation after the above dlopen(3DL) has occurred, the runtime linkers
normal search for the relocation symbol is to look in prog and then A.so.1. The
runtime linker does not follow through and look in B.so.1 or C.so.1.
These symbol search and symbol visibility attributes maintain associations between
objects based on their introduction into the process address space, and on any
dependency relationship between the objects. Assigning the objects associated with a
given dlopen(3DL) to a unique group ensures that only objects associated with the
same dlopen(3DL) are allowed to look up symbols within themselves and their
related dependencies.
and the prog application used dlopen(3DL) to load this shared object in addition to
the shared object B.so.1. The following figure illustrates the symbol lookup
releationship between the objects.
B.so.1 C.so.1
prog A.so.1
D.so.1 E.so.1
Suppose that both B.so.1 and D.so.1 contain a definition for the symbol foo, and
both C.so.1 and E.so.1 contain a relocation that requires this symbol. Because of
the association of objects to a unique group, C.so.1 is bound to the definition in
B.so.1, and E.so.1 is bound to the definition in D.so.1. This mechanism is
intended to provide the most intuitive binding of objects obtained from multiple calls
to dlopen(3DL).
In the following example, the shared objects O.so.1 and P.so.1 have the same
common dependency.
$ ldd O.so.1
Z.so.1 => ./Z.so.1
$ ldd P.so.1
Z.so.1 => ./Z.so.1
In this example, the prog application will dlopen(3DL) each of these shared objects.
Because the shared object Z.so.1 is a common dependency of both O.so.1 and
P.so.1, Z.so.1 is assigned to both of the groups that are associated with the two
dlopen(3DL) calls. This relationship is shown in the following figure.
O.so.1
P.so.1
Z.so.1 is available for both O.so.1 and P.so.1 to look up symbols. More
importantly, as far as dlopen(3DL) ordering is concerned, Z.so.1 is also be able to
look up symbols in both O.so.1 and P.so.1.
Therefore, if both O.so.1 and P.so.1 contain a definition for the symbol foo, which
is required for a Z.so.1 relocation, the actual binding that occurs is unpredictable
because it is affected by the order of the dlopen(3DL) calls. If the functionality of
symbol foo differs between the two shared objects in which it is defined, the overall
outcome of executing code within Z.so.1 might vary depending on the applications
dlopen(3DL) ordering.
Isolating a Group
The default assignment of world symbol search scope to the objects obtained by a
dlopen(3DL) can be reduced to group by augmenting the mode argument with the
RTLD_GROUP flag. Under this mode, any objects obtained through a dlopen(3DL)
will only be allowed to look for symbols within their own group.
Using the link-editors -B group option, you can assign the group symbol search
scope to objects when they are built.
Object Hierarchies
If an initial object, obtained from a dlopen(3DL), was to use dlopen(3DL) to open a
secondary object, both objects would be assigned to a unique group. This situation can
prevent either object from locating symbols from one another.
In some implementations the initial object has to export symbols for the relocation of
the secondary object. This requirement can be satisfied by one of two mechanisms:
Making the initial object an explicit dependency of the second object
Use the RTLD_PARENT mode flag to dlopen(3DL) the secondary object
If the initial object is an explicit dependency of the secondary object, the initial object
is assigned to the secondary objects group. The initial object is therefore able to
provide symbols for the secondary objects relocation.
There is one small difference between these two techniques. If you specify an explicit
dependency, the dependency itself becomes part of the secondary objects
dlopen(3DL) dependency tree, and thus becomes available for symbol lookup with
dlsym(3DL). If you obtain the secondary object with RTLD_PARENT, the initial object
does not become available for symbol lookup with dlsym(3DL).
main()
{
void * handle;
int * dptr, (* fptr)();
return ((*fptr)(*dptr));
}
The symbols foo and bar are searched for in the file foo.so.1, followed by any
dependencies that are associated with this file. The function foo is then called with
the single argument bar as part of the return() statement.
The application prog, built using the above file main.c, contains the following
dependencies:
$ ldd prog
libdl.so.1 => /usr/lib/libdl.so.1
libc.so.1 => /usr/lib/libc.so.1
If the file name specified in the dlopen(3DL) had the value 0, the symbols foo and
bar are searched for in prog, followed by /usr/lib/libdl.so.1, and finally
/usr/lib/libc.so.1.
Once the handle has indicated the root at which to start a symbol search, the search
mechanism follows the same model as described in Symbol Lookup on page 68.
If the required symbol cannot be located, dlsym(3DL) returns a NULL value. In this
case, dlerror(3DL) can be used to indicate the true reason for the failure. In the
following example, the application prog is unable to locate the symbol bar.
$ prog
dlsym: ld.so.1: main: fatal: bar: cant find symbol
This mechanism provides a robust and flexible alternative to the use of undefined
weak references, discussed in Weak Symbols on page 45.
Using Interposition
The special handle RTLD_NEXT enables an application to locate the next symbol in a
symbol scope. For example, if the application prog contained the following code
fragment:
if ((fptr = (int (*)())dlsym(RTLD_NEXT, "foo")) == NULL) {
(void) printf("dlsym: %s\n", dlerror());
exit (1);
}
return ((*fptr)());
then foo is searched for in the shared objects associated with prog, which in this case
are /usr/lib/libdl.so.1 and then /usr/lib/libc.so.1. If this code fragment
was contained in the file B.so.1 from the example shown in Figure 31, then foo is
searched for in the associated shared object C.so.1 only.
void *
malloc(size_t size)
{
static void * (* fptr)() = 0;
char buffer[50];
This shared object can be interposed before the system library /usr/lib/libc.so.1
where malloc(3C) usually resides. Any calls to malloc() are now interposed upon
before the original function is called to complete the allocation:
$ cc -o malloc.so.1 -G -K pic malloc.c
$ cc -o prog file1.o file2.o ..... -R. malloc.so.1
$ prog
malloc: 0x32 bytes
malloc: 0x14 bytes
..........
Note Users of any interposition technique must be careful to handle any possibility of
recursion. The previous example formats the diagnostic message using sprintf(3C),
instead of using printf(3C) directly, to avoid any recursion caused by printf(3C)s
possible use of malloc(3C).
Feature Checking
Dynamic objects built by the link-editor sometimes require new runtime linker
features. The function _check_rtld_feature() can be used to check if the runtime
features required for execution are supported by the running runtime linker. The
runtime features currently identified are listed in Table 746.
Debugging Library
This debugging library helps you understand, or debug, the execution of applications
and dependencies. Although the type of information displayed using this library is
expected to remain constant, the exact format of the information might change slightly
from release to release.
Some of the debugging output might be unfamiliar to those who do not have an
intimate knowledge of the runtime linker. However, many aspects may be of general
interest to you.
The tokens available with this debugging option can be displayed by using
LD_DEBUG=help. Any dynamic executable can be used to solicit this information, as
the process itself terminates following the display of the information. For example:
$ LD_DEBUG=help prog
11693:
11693: For debugging the runtime linking of an application:
11693: LD_DEBUG=token1,token2 prog
11693: enables diagnostics to the stderr. The additional
11693: option:
11693: LD_DEBUG_OUTPUT=file
11693: redirects the diagnostics to an output file created
11593: using the specified name and the process id as a
11693: suffix. All diagnostics are prepended with the
11693: process id.
11693:
11693:
11693: audit display runtime link-audit processing
11693: basic provide basic trace information/warnings
11693: bindings display symbol binding; detail flag shows
11693: absolute:relative addresses
11693: detail provide more information in conjunction with other
11693: options
11693: files display input file processing (files and libraries)
11693: help display this help message
This example shows the options meaningful to the runtime linker. The exact options
might differ from release to release.
One of the most useful debugging options is to display the symbol bindings that occur
at runtime. The following example uses a very trivial dynamic executable that has a
dependency on two local shared objects.
$ cat bar.c
int bar = 10;
$ cc -o bar.so.1 -K pic -G bar.c
$ cat foo.c
foo(int data)
{
return (data);
}
$ cc -o foo.so.1 -K pic -G foo.c
$ cat main.c
extern int foo();
extern int bar;
main()
{
return (foo(bar));
}
$ cc -o prog main.c -R/tmp:. foo.so.1 bar.so.1
When the runtime linker performs a function relocation, it rewrites data associated
with the functions .plt so that any subsequent calls will go directly to the function.
The environment variable LD_BIND_NOT can be set to any value to prevent this data
update. By using this variable together with the debugging request for detailed
bindings, you can get a complete runtime account of all function binding. The output
from this combination can be excessive, in which case the performance of the
application is degraded.
You can use LD_DEBUG to display the various search paths used. For example, the
search path mechanism used to locate any dependencies can be displayed by setting
LD_DEBUG=libs.
$ LD_DEBUG=libs prog
11775:
11775: find object=foo.so.1; searching
11775: search path=/tmp:. (RPATH from file prog)
11775: trying path=/tmp/foo.so.1
11775: trying path=./foo.so.1
11775:
11775: find object=bar.so.1; searching
11775: search path=/tmp:. (RPATH from file prog)
11775: trying path=/tmp/bar.so.1
11775: trying path=./bar.so.1
11775: .......
The runpath recorded in the application prog affects the search for the two
dependencies foo.so.1 and bar.so.1.
In a similar manner, the search paths of each symbol lookup can be displayed by
setting LD_DEBUG=symbols. If this is combined with a bindings request, you can
obtain a complete picture of the symbol relocation process.
$ LD_DEBUG=bindings,symbols
11782: .......
11782: symbol=bar; lookup in file=./foo.so.1 [ ELF ]
11782: symbol=bar; lookup in file=./bar.so.1 [ ELF ]
11782: binding file=prog to file=./bar.so.1: symbol bar
11782: .......
11782: transferring control: prog
11782: .......
11782: symbol=foo; lookup in file=prog [ ELF ]
11782: symbol=foo; lookup in file=./foo.so.1 [ ELF ]
11782: binding file=prog to file=./foo.so.1: symbol foo
11782: .......
Debugger Module
The debugger module provides a set of dcmds and walkers that can be loaded under
mdb(1). This module can be used to inspect various internal data structures of the
runtime linker. Much of this information requires familiarity with the internals of the
runtime linker, and may change from release to release. However, some elements of
these data structures reveal the basic components of a dynamically linked process and
may aid general debugging.
The following examples show some simple scenarios of using mdb(1) with the runtime
linker debugger module.
$ cat main.c
#include <dlfnc.h>
int main()
{
void * handle;
void (* fptr)();
(*fptr)();
return (0);
}
$ cc -o main main.c -R. -ldl
If mdb(1) has not automatically loaded the debugger module, ld.so, explicitly do so.
The capabilities of the debugger module can then be inspected.
$ mdb main
> ::load ld.so
> ::dmods -l ld.so
ld.so
-----------------------------------------------------------------
dcmd Bind - Display a Binding descriptor
dcmd Callers - Display Rt_map CALLERS binding descriptors
dcmd Depends - Display Rt_map DEPENDS binding descriptors
dcmd ElfDyn - Display Elf_Dyn entry
dcmd ElfEhdr - Display Elf_Ehdr entry
dcmd ElfPhdr - Display Elf_Phdr entry
dcmd Groups - Display Rt_map GROUPS group handles
The objects .dynamic section can be displayed with the ElfDyn dcmd. The
following example shows the first 4 entries.
> 0x000207bc,4::ElfDyn
Elf_Dyn located at: 0x207bc
0x207bc NEEDED 0x0000010f
Elf_Dyn located at: 0x207c4
0x207c4 NEEDED 0x00000124
Elf_Dyn located at: 0x207cc
0x207cc INIT 0x00010710
mdb(1) is also very useful for setting deferred break points. In this example, it might be
useful to put a break point on the function foo(). However, until the dlopen(3DL) of
foo.so.1 occurs, this symbol isnt known to the debugger. A deferred break point
instructs the debugger to set a real breakpoint when the dynamic object is loaded.
> ::bp foo.so.1foo
> :c
> mdb: Youve got symbols!
> mdb: stop at foo.so.1foo
mdb: target stopped at:
foo.so.1foo: save %sp, -0x68, %sp
The link-map for foo.so.1 shows the handle returned by dlopen(3DL). You can
expand this structure using Handles.
> 0xff3f9ca4::Handles -v
HANDLES for ./foo.so.1
----------------------------------------------
HANDLE: 0xff3f9f60 Alist[used 1: total 1]
----------------------------------------------
Group Handle located at: 0xff3f9f28
----------------------------------------------
owner: ./foo.so.1
flags: 0x00000000 [ 0 ]
refcnt: 1 depends: 0xff3f9fa0 Alist[used 2: total 4]
----------------------------------------------
Group Descriptor located at: 0xff3f9fac
depend: 0xff3f9ca4 ./foo.so.1
flags: 0x00000003 [ AVAIL-TO-DLSYM,ADD-DEPENDENCIES ]
----------------------------------------------
Group Descriptor located at: 0xff3f9fd8
depend: 0xff37006c ./bar.so.1
flags: 0x00000003 [ AVAIL-TO-DLSYM,ADD-DEPENDENCIES ]
The dependencies of a handle are a list of link-maps that represent the objects of the
handle that can satisfy a dlsym(3DL) request. In this case, the dependencies are
foo.so.1 and bar.so.1.
Shared Objects
Shared objects are one form of output created by the link-editor and are generated by
specifying the -G option. In the following example, the shared object libfoo.so.1 is
generated from the input file foo.c.
$ cc -o libfoo.so.1 -G -K pic foo.c
A shared object is an indivisible unit that is generated from one or more relocatable
objects. Shared objects can be bound with dynamic executables to form a runable
process. As their name implies, shared objects can be shared by more than one
application. Because of this potentially far-reaching effect, this chapter describes this
form of link-editor output in greater depth than has been covered in previous
chapters.
Any input shared objects become dependencies of this output file. A small amount of
bookkeeping information is maintained within the output file to describe these
dependencies. The runtime linker interprets this information and completes the
processing of these shared objects as part of creating a runable process.
The following sections expand upon the use of shared objects within the compilation
and runtime environments. These environments are introduced in Runtime Linking
on page 21.
101
Naming Conventions
Neither the link-editor nor the runtime linker interprets any file by virtue of its file
name. All files are inspected to determine their ELF type (see ELF Header
on page 180). This information enables the link-editor to deduce the processing
requirements of the file. However, shared objects usually follow one of two naming
conventions, depending on whether they are being used as part of the compilation
environment or the runtime environment.
When used as part of the compilation environment, shared objects are read and
processed by the link-editor. Although these shared objects can be specified by explicit
file names as part of the command passed to the link-editor, the -l option is usually
used to take advantage of the link-editors library search capabilities. See Shared
Object Processing on page 30.
When used as part of the runtime environment, shared objects are read and processed
by the runtime linker. To allow for change in the exported interface of the shared
object over a series of software releases, provide the shared object as a versioned file
name.
A versioned file name commonly takes the form of a .so suffix followed by a version
number. For example, /usr/lib/libc.so.1 is the shared object representation of
version one of the standard C library made available to the runtime environment.
If a shared object is never intended for use within a compilation environment, its
name might drop the conventional lib prefix. Examples of shared objects that fall into
this category are those used solely with dlopen(3DL). A suffix of .so is still
recommended to indicate the actual file type. In additions, a version number is
strongly recommended to provide for the correct binding of the shared object across a
series of software releases. Chapter 5 describes versioning in more detail.
During the link-edit of a shared object, its runtime name can be recorded within the
shared object itself by using the -h option. In the following example, the shared
objects runtime name libfoo.so.1, is recorded within the file itself. This
identification is known as an soname.
$ cc -o ../tmp/libfoo.so -G -K pic -h libfoo.so.1 foo.c
The following example shows how the soname recording can be displayed using
dump(1) and referring to the entry that has the SONAME tag.
$ dump -Lvp ../tmp/libfoo.so
../tmp/libfoo.so:
[INDEX] Tag Value
[1] SONAME libfoo.so.1
.........
When the link-editor processes a shared object that contains an soname, this is the
name that is recorded as a dependency within the output file being generated.
If this new version of libfoo.so is used during the creation of the dynamic
executable prog from the previous example, all three methods of creating the
executable result in the same dependency recording.
In the previous examples, the -h option is used to specify a simple file name, that has
no / in the name. This convention enables the runtime linker to use a set of rules to
locate the actual file. See Locating Shared Object Dependencies on page 64 for more
details.
An archive can be built from one or more shared objects and then used to generate a
dynamic executable or shared object. Shared objects can be extracted from the archive
to satisfy the requirements of the link-edit. Unlike the processing of relocatable objects,
which are concatenated to the output file being created, any shared objects extracted
from the archive will be recorded as dependencies. See Archive Processing
on page 29 for more details on the criteria for archive extraction.
Because a file with this concatenated name is unlikely to exist at runtime, providing an
soname within the shared object is the only means of generating a meaningful runtime
file name for the dependency.
Note The runtime linker does not extract objects from archives. Therefore, in the
previous example, the required shared object dependencies must be extracted from the
archive and made available to the runtime environment.
Conflicts in dependency names can occur if two shared objects used as input files to a
link-edit both contain the same soname. For example:
$ cc -o libfoo.so -G -K pic -h libsame.so.1 foo.c
$ cc -o libbar.so -G -K pic -h libsame.so.1 bar.c
$ cc -o prog main.o -L. -lfoo -lbar
ld: fatal: recording name conflict: file ./libfoo.so and \
file ./libbar.so provide identical dependency names: libsame.so.1
ld: fatal: File processing errors. No output written to prog
A similar error condition occurs if the file name of a shared object that does not have a
recorded soname matches the soname of another shared object used during the same
link-edit.
If the runtime name of a shared object being generated matches one of its
dependencies, the link-editor also reports a name conflict
$ cc -o libbar.so -G -K pic -h libsame.so.1 bar.c -L. -lfoo
ld: fatal: recording name conflict: file ./libfoo.so and \
-h option provide identical dependency names: libsame.so.1
ld: fatal: File processing errors. No output written to libbar.so
libfoo.so:
The shared object is responsible for specifying any runpath required to locate its
dependencies. Any runpath specified in the dynamic executable is only used to locate
the dependencies of the dynamic executable. These runpaths are not used to locate
any dependencies of the shared objects.
The environment variable LD_LIBRARY_PATH has a more global scope. Any path
names specified using this variable are used by the runtime linker to search for any
shared object dependencies. Although useful as a temporary mechanism that
influences the runtime linkers search path, the use of this environment variable is
strongly discouraged in production software. See Directories Searched by the
Runtime Linker on page 64 for a more extensive discussion.
Dependency Ordering
When dynamic executables and shared objects have dependencies on the same
common shared objects, the order in which the objects are processed can become less
predictable.
For example, assume a shared object developer generates libfoo.so.1 with the
following dependencies:
$ ldd libfoo.so.1
libA.so.1 => ./libA.so.1
libB.so.1 => ./libB.so.1
libC.so.1 => ./libC.so.1
If you create a dynamic executable prog, using this shared object, and define an
explicit dependency on libC.so.1, the resulting shared object order will be:
$ cc -o prog main.c -R. -L. -lC -lfoo
$ ldd prog
libC.so.1 => ./libC.so.1
libfoo.so.1 => ./libfoo.so.1
libA.so.1 => ./libA.so.1
libB.so.1 => ./libB.so.1
Interfaces are defined to act as standard filters by using the link-editors -F flag.
This flag is qualified with the name of one or more filtees that must supply the
symbol definition at runtime.
Interfaces are defined to act as auxiliary filters by using the link-editors -f flag.
This flag is qualified with the name of one or more filtees that may supply the
symbol definition at runtime.
char * foo()
{
return("defined in filtee");
}
$ cc -o filtee.so.1 -G -K pic filtee.c
To declare all the interfaces offered by a shared object to be filters, the shared object is
defined a filter by using the link-editors -F flag.
char * foo()
{
return (0);
}
$ LD_OPTIONS=-F filtee.so.1 \
cc -o filter.so.1 -G -K pic -h filter.so.1 -R. filter.c
$ elfdump -d filter.so.1 | egrep "SONAME|FILTER"
[2] SONAME 0xee filter.so.1
[3] FILTER 0xfb filtee.so.1
For example, the following dynamic executable prog, references the symbols foo and
bar, which are resolved during link-edit from the filter filter.so.1. The execution
of prog results in foo and bar being obtained from the filtee filtee.so.1, not from
the filter filter.so.1.
$ cat main.c
extern char * bar, * foo();
main()
In these examples, the filtee filtee.so.1 is uniquely associated to the filter, and is
not available to satisfy symbol lookup from any other objects that might be loaded as a
consequence of executing prog.
The /usr/lib/libdl.so.1 filter defines the user interface to the runtime linker.
This interface provides an abstraction between the symbols referenced in a
compilation environment from libdl.so.1 and the actual implementation binding
produced within the runtime environment from ld.so.1(1).
Because the code in a standard filter is never referenced at runtime, adding content to
any functions defined as filters is redundant. Any filter code might require relocation,
which would result in an unnecessary overhead when processing the filter at runtime.
Functions are best defined as empty routines, or directly from a mapfile. See
Defining Additional Symbols on page 46.
Note The link-editor uses the ELF class of the first relocatable file that is processed to
govern the class of object that is created. Use the link-editors -64 option to create a
64bit filter solely from a mapfile.
When generating data symbols within a filter, always initialize the data items. The
resulting data definition insures that references are established correctly from dynamic
executables.
To declare all of the interfaces offered by a shared object to be auxiliary filters, the
shared object is defined to be an auxiliary filter using the link-editors -f flag.
char * foo()
{
return ("defined in filter");
}
$ LD_OPTIONS=-f libbar.so.1 \
cc -o filter.so.1 -G -K pic -h filter.so.1 -R. filter.c
$ elfdump -d filter.so.1 | egrep "SONAME|AUXILIARY"
[2] SONAME 0xee filter.so.1
[3] AUXILIARY 0xfb filtee.so.1
main()
{
(void) printf("foo is %s: bar is %s\n", foo(), bar);
}
$ cc -o prog main.c -R. filter.so.1
$ prog
foo is defined in filtee: bar is defined in filter
In these examples, the filtee filtee.so.1 is uniquely associated to the filter, and is
not available to satisfy symbol lookup from any other objects that might be loaded as a
consequence of executing prog.
Note The environment variable LD_NOAUXFLTR can be set to disable the runtime
linkers auxiliary filter processing. Because auxiliary filters are frequently employed to
provide platform specific optimizations, this option can be useful in evaluating filtee
use and their performance impact.
Filtee Processing
The runtime linkers processing of a filter defers the loading of a filtee until a filter
symbol is referenced. This implementation is analogous to the filter performing a
dlopen(3DL), using mode RTLD_LOCAL, on each of its filtees as they are required.
This implementation accounts for differences in dependency reporting that can be
produced by tools such as ldd(1).
The link-editors -z loadfltr option can be used when creating a filter to cause the
immediate processing of its filtees at runtime. In addition, the immediate processing of
any filtees within a process can be triggered by setting the LD_LOADFLTR environment
variable to any value.
Although the actual code within a shared object will directly affect the performance of
a running process, the performance issues focused upon here target the runtime
processing of the shared object itself. The following sections investigate this processing
in more detail by looking at aspects such as text size and purity, together with
relocation overhead.
Analyzing Files
Various tools are available to analyze the contents of an ELF file. To display the size of
a file use the size(1) command. For example:
$ size -x libfoo.so.1
59c + 10c + 20 = 0x6c8
The first example indicates the size of the shared objects text, data, and bss, a
categorization used in previous releases of the SunOS operating system.
The ELF format provides a finer granularity for expressing data within a file by
organizing the data into sections. The second example displays the size of each of the
files loadable sections.
Sections are allocated to units known as segments, some of which describe how
portions of a file are mapped into memory (see the mmap(2) man page). These loadable
segments can be displayed by using the dump(1) command and examining the LOAD
entries. For example:
$ dump -ov libfoo.so.1
libfoo.so.1:
***** PROGRAM EXECUTION HEADER *****
Type Offset Vaddr Paddr
Filesz Memsz Flags Align
There are two loadable segments in the shared object libfoo.so.1, commonly
referred to as the text and data segments. The text segment is mapped to allow reading
and execution of its contents (r-x), whereas the data segment is mapped to also allow
its contents to be modified (rwx). The memory size (Memsz) of the data segment
differs from the file size (Filesz). This difference accounts for the .bss section,
which is part of the data segment, and is dynamically created when the segment is
loaded.
Programmers usually think of a file in terms of the symbols that define the functions
and data elements within their code. These symbols can be displayed using nm(1). For
example:
$ nm -x libfoo.so.1
The section that contains a symbol can be determined by referencing the section index
(Shndx) field from the symbol table and by using dump(1) to display the sections
within the file. For example:
$ dump -hv libfoo.so.1
libfoo.so.1:
**** SECTION HEADER TABLE ****
[No] Type Flags Addr Offset Size Name
.........
[7] PBIT -AI 0x538 0x538 0x1c .init
The output from both the previous nm(1) and dump(1) examples shows the association
of the functions _init, foo, and _fini to the sections .init, .text and .fini.
These sections, because of their read-only nature, are part of the text segment.
Similarly, the data arrays data, and bss are associated with the sections .data and
.bss respectively. These sections, because of their writable nature, are part of the data
segment.
Underlying System
When an application is built using a shared object, the entire loadable contents of the
object are mapped into the virtual address space of that process at runtime. Each
process that uses a shared object starts by referencing a single copy of the shared
object in memory.
Relocations within the shared object are processed to bind symbolic references to their
appropriate definitions. This results in the calculation of true virtual addresses that
could not be derived at the time the shared object was generated by the link-editor.
These relocations usually result in updates to entries within the processs data
segments.
The memory management scheme underlying the dynamic linking of shared objects
shares memory among processes at the granularity of a page. Memory pages can be
shared as long as they are not modified at runtime. If a process writes to a page of a
shared object when writing a data item, or relocating a reference to a shared object, it
generates a private copy of that page. This private copy will have no effect on other
users of the shared object. However, this page has lost any benefit of sharing between
other processes. Text pages that become modified in this manner are referred to as
impure.
The segments of a shared object that are mapped into memory fall into two basic
categories; the text segment, which is read-only, and the data segment, which is
read-write. See Analyzing Files on page 112 on how to obtain this information from
an ELF file. An overriding goal when developing a shared object is to maximize the
text segment and minimize the data segment. This optimizes the amount of code
sharing while reducing the amount of processing needed to initialize and use a shared
object. The following sections present mechanisms that can help achieve this goal.
For small applications a typical thread of execution may reference all the applications
dependencies. The application loads all of its dependencies whether they are defined
lazy loadable or not. However, under lazy loading, dependency processing may be
deferred from process startup and spread throughout the processs execution.
Position-Independent Code
The compiler can generate position-independent code under the -K pic option.
Whereas the code within a dynamic executable is usually tied to a fixed address in
memory, position-independent code can be loaded anywhere in the address space of a
process. Because the code is not tied to a specific address, it will execute correctly
without page modification at a different address in each process that uses it. This code
creates programs that require the smallest amount of page modification at runtime.
If a shared object is built from code that is not position-independent, the text segment
will usually require a large number of relocations to be performed at runtime.
Although the runtime linker is equipped to handle this, the system overhead this
creates can cause serious performance degradation.
You can identify a shared object that requires relocations against its text segment. Use
dump(1) and inspect the output for any TEXTREL entry. For example:
$ cc -o libfoo.so.1 -G -R. foo.c
$ dump -Lv libfoo.so.1 | grep TEXTREL
[9] TEXTREL 0
Note The value of the TEXTREL entry is irrelevant. Its presence in a shared object
indicates that text relocations exist.
To prevent the creation of a shared object that contains text relocations use the
link-editors -z text flag. This flag causes the link-editor to generate diagnostics
indicating the source of any non-position-independent code used as input. Such code
results in a failure to generate the intended shared object. For example:
$ cc -o libfoo.so.1 -z text -G -R. foo.c
Text relocation remains referenced
against symbol offset in file
foo 0x0 foo.o
bar 0x8 foo.o
ld: fatal: relocations remain against allocatable but \
Two relocations are generated against the text segment because of the
non-position-independent code generated from the file foo.o. Where possible, these
diagnostics indicate any symbolic references that are required to carry out the
relocations. In this case, the relocations are against the symbols foo and bar.
Another common cause of creating text relocations when generating a shared object is
by including hand-written assembler code that has not been coded with the
appropriate position-independent prototypes.
Note You may want to experiment with some simple source files to determine coding
sequences that enable position-independence. Use the compilers ability to generate
intermediate assembler output.
The global offset table is an array of pointers, the size of whose entries are constant for
32bit (4bytes) and 64bit (8bytes). The following code sequence makes reference to
an entry under -K pic:
ld [%l7 + j], %o0 ! load &j into %o0
This code sequence provides a 13bit displacement constant for the global offset table
entry. This displacement therefore provides for 2048 unique entries for 32bit objects,
and 1024 unique entries for 64bit objects. If an object is built that requires more than
the available number of entries, the link-editor produces a fatal error:
$ cc -K pic -G -o lobfoo.so.1 a.o b.o ... z.o
ld: fatal: too many symbols require small PIC references:
have 2050, maximum 2048 -- recompile some modules -K PIC.
To overcome this error condition, compile some of the input relocatable objects with
the -K PIC option. This option provides a 32bit constant for the global offset table
entry:
sethi %hi(j), %g1
or %g1, %lo(j), %g1 ! get 32-bit constant GOT offset
ld [%l7 + %g1], %o0 ! load &j into %o0
Ideally, frequently accessed data items benefit from using the -K pic model. You can
reference a single entry using both models. However, determining which relocatable
objects should be compiled with either option can be time consuming, and the
performance improvement realized small. A recompilation of all relocatable objects
with the -K PIC option is typically easier.
Unused sections are displayed during a link-edit when using the link-editors
debugging token -D unused. Sections identified as unused should be removed from
the link-edit, or eliminated using the link-editors -z ignore option.
You can improve the link-editors ability to eliminate sections by defining the shared
objects external interfaces. By defining an interface, global symbols that are not
defined as part of the interface are reduced to locals. These reduced symbols, if
unreferenced from other objects, are now clearly identified as candidates for
elimination.
Individual functions and data variables can be eliminated by the link-editor if these
items are assigned to their own sections. This section refinement is achieved using
compiler options such as -xF. Earlier compilers only provided for the assignment of
functions to their own sections. Newer compilers have extended the -xF syntax to
assign data variables to their own sections. Earlier compilers required C++ exception
handling to be disabled when using -xF. This restriction has been dropped with later
compilers.
If all allocatable sections from a relocatable object can be eliminated, the entire file is
discarded from the link-edit.
Maximizing Shareability
As mentioned in Underlying System on page 114, only a shared objects text
segment is shared by all processes that use it. The objects data segment typically is
not shared. Each process using a shared object, generates a private memory copy of its
entire data segment as data items within the segment are written to. Reduce the data
segment, either by moving data elements that are never written to the text segment, or
by removing the data items completely.
The following sections describe several mechanisms that can be used to reduce the
size of the data segment.
In contrast, the following character string resides in the .rodata section, which is the
read-only data section contained within the text segment:
const char * rdstr = "this is a read-only string";
Reducing the data segment by moving read-only elements into the text segment is
admirable. However, moving data elements that require relocations can be
counterproductive. For example, examine the following array of strings:
char * rdstrs[] = { "this is a read-only string",
"this is another read-only string" };
This definition ensures that the strings and the array of pointers to these strings are
placed in a .rodata section. Unfortunately, although the user perceives the array of
addresses as read-only, these addresses must be relocated at runtime. This definition
therefore results in the creation of text relocations. Representing it as:
const char * rdstrs[] = { ..... };
insures the array pointers are maintained in the writable data segment where they can
be relocated. The array strings are maintained in the read-only text segment.
foo()
{
......
(void) fprintf(stderr, Errmsg, error);
......
The main candidates for this sort of data reduction are strings. String usage in a
shared object can be investigated using strings(1). The following example generates
a sorted list of the data strings within the file libfoo.so.1. Each entry in the list is
prefixed with the number of occurrences of the string.
$ strings -10 libfoo.so.1 | sort | uniq -c | sort -rn
Organizing frequently used routines and their data to an adjacent set of pages
frequently improves performance because it improves the locality of reference. When a
process calls one of these functions, the function might already be in memory because
of its proximity to the other frequently used functions. Similarly, grouping interrelated
functions improves locality of references. For example, if every call to the function
foo() results in a call to the function bar(), place these functions on the same page.
Tools like cflow(1), tcov(1), prof(1) and gprof(1) are useful in determining code
coverage and profiling.
Isolate related functionality to its own shared object. The standard C library has
historically been built containing many unrelated functions. Only rarely, for example,
will any single executable use everything in this library. Because of widespread use,
determining what set of functions are really the most frequently used is also
somewhat difficult. In contrast, when designing a shared object from scratch, maintain
only related functions within the shared object. This will improve locality of reference
and has the side effect of reducing the objects overall size.
Relocations
In Relocation Processing on page 67, the mechanisms by which the runtime linker
relocates dynamic executables and shared objects to create a runable process was
covered. Symbol Lookup on page 68 and When Relocations Are Performed
on page 71 categorized this relocation processing into two areas to simplify and help
illustrate the mechanisms involved. These same two categorizations are also ideally
suited for considering the performance impact of relocations.
Symbol Lookup
When the runtime linker needs to look up a symbol, by default it does so by searching
in each object. The runtime linker starts with the dynamic executable, and progresses
through each shared object in the same order that the objects are loaded. In many
instances, the shared object that requires a symbolic relocation turns out to be the
provider of the symbol definition.
In this situation, if the symbol used for this relocation is not required as part of the
shared objects interface, then this symbol is a strong candidate for conversion to a
static or automatic variable. A symbol reduction can also be applied to removed
The only global data items that should be visible from a shared object are those that
contribute to its user interface. Historically this has been a hard goal to accomplish,
because global data are often defined to allow reference from two or more functions
located in different source files. By applying symbol reduction, unnecessary global
symbols can be removed. See Reducing Symbol Scope on page 52. Any reduction in
the number of global symbols exported from a shared object results in lower relocation
costs and an overall performance improvement.
The use of direct bindings can also significantly reduce the symbol lookup overhead
within a dynamic process that has many symbolic relocations and many
dependencies. See Direct Binding on page 69.
Initialization relocation costs can also be deferred by converting data references into
function references. For example, you can return data items by a functional interface.
This conversion usually results in a perceived performance improvement because the
initialization relocation costs are effectively spread throughout the processs execution.
Some of the functional interfaces might never be called by a particular invocation of a
process, thus removing their relocation overhead altogether.
The advantage of using a functional interface can be seen in the section, Copy
Relocations on page 122. This section examines a special, and somewhat expensive,
relocation mechanism employed between dynamic executables and shared objects. It
also provides an example of how this relocation overhead can be avoided.
Copy Relocations
Shared objects are usually built with position-independent code. References to
external data items from code of this type employs indirect addressing through a set of
tables. See Position-Independent Code on page 115 for more details. These tables are
updated at runtime with the real address of the data items. These updated tables
enable access to the data without the code itself being modified.
Because the symbol assigned to this space is global, it is used to satisfy any references
from any shared objects. The dynamic executable inherits the data item. Any other
objects within the process that make reference to this item are bound to this copy. The
original data from which the copy is made effectively becomes unused.
The following example of this mechanism uses an array of system error messages that
is maintained within the standard C library. In previous SunOS operating system
releases, the interface to this information was provided by two global variables,
sys_errlist[], and sys_nerr. The first variable provided the array of error
message strings, while the second conveyed the size of the array itself. These variables
were commonly used within an application in the following manner:
$ cat foo.c
extern int sys_nerr;
extern char * sys_errlist[];
char *
error(int errnumb)
{
if ((errnumb < 0) || (errnumb >= sys_nerr))
return (0);
return (sys_errlist[errnumb]);
The application uses the function error to provide a focal point to obtain the system
error message associated with the number errnumb.
Examining a dynamic executable built using this code shows the implementation of
the copy relocation in more detail:
$ cc -o prog main.c foo.c
$ nm -x prog | grep sys_
[36] |0x00020910|0x00000260|OBJT |WEAK |0x0 |16 |sys_errlist
[37] |0x0002090c|0x00000004|OBJT |WEAK |0x0 |16 |sys_nerr
$ dump -hv prog | grep bss
[16] NOBI WA- 0x20908 0x908 0x268 .bss
$ dump -rv prog
.rela.bss:
Offset Symndx Type Addend
The link-editor has allocated space in the dynamic executables .bss to receive the
data represented by sys_errlist and sys_nerr. These data are copied from the C
library by the runtime linker at process initialization. Thus, each application that uses
these data gets a private copy of the data in its own data segment.
There are two drawbacks to this technique. First, each application pays a performance
penalty for the overhead of copying the data at runtime. Second, the size of the data
array sys_errlist has now become part of the C librarys interface. Suppose the
size of this array were to change, prehaps as new error messages are added. Any
dynamic executables that reference this array have to undergo a new link-edit to be
able to access any of the new error messages. Without this new link-edit, the allocated
space within the dynamic executable is insufficient to hold the new data.
These drawbacks can be eliminated if the data required by a dynamic executable are
provided by a functional interface. The ANSI C function strerror(3C) returns a
pointer to the appropriate error string, based on the error number supplied to it. One
implementation of this function might be:
$ cat strerror.c
static const char * sys_errlist[] = {
"Error 0",
"Not owner",
"No such file or directory",
......
};
static const int sys_nerr =
sizeof (sys_errlist) / sizeof (char *);
The error routine in foo.c can now be simplified to use this functional interface. This
simplification in turn removes any need to perform the original copy relocations at
process initialization.
Additionally, because the data are now local to the shared object, the data are no
longer part of its interface. The shared object therefore has the flexibility of changing
the data without adversely effecting any dynamic executables that use it. Eliminating
data items from a shared objects interface generally improves performance while
making the shared objects interface and code easier to maintain.
ldd(1), when used with either the -d or -r options, can verify any copy relocations
that exist within a dynamic executable.
For example, suppose the dynamic executable prog had originally been built against
the shared object libfoo.so.1 and the following two copy relocations had been
recorded:
$ nm -x prog | grep _size_
[36] |0x000207d8|0x40|OBJT |GLOB |15 |_size_gets_smaller
[39] |0x00020818|0x40|OBJT |GLOB |15 |_size_gets_larger
$ dump -rv size | grep _size_
0x207d8 _size_gets_smaller R_SPARC_COPY 0
0x20818 _size_gets_larger R_SPARC_COPY 0
A new version of this shared object is supplied that contains different data sizes for
these symbols:
$ nm -x libfoo.so.1 | grep _size_
[26] |0x00010378|0x10|OBJT |GLOB |8 |_size_gets_smaller
[28] |0x00010388|0x80|OBJT |GLOB |8 |_size_gets_larger
ldd(1) shows that the dynamic executable will copy as much data as the shared object
has to offer, but only accepts as much as its allocated space allows.
Using -B symbolic
The link-editors -B symbolic option enables you to bind symbol references to their
global definitions within a shared object. This option is historic, in that it was designed
for use in creating the runtime linker itself.
If a symbolically bound symbol is interposed upon, then references to the symbol from
outside of the symbolically bound object bind to the interposer. The object itself is
already bound internally. Essentially, two symbols with the same name are now being
referenced from within the process. A symbolically bound data symbol that results in a
copy relocation creates the same interposition situation. See Copy Relocations
on page 122.
Note Symbolically bound shared objects are identified by the .dynamic flag
DF_SYMBOLIC. This flag is informational only. The runtime linker processes symbol
lookups from these objects in the same manner as any other object. Any symbolic
binding is assumed to have been created at the link-edit phase.
When profiling is enabled, a profile data file is created, if it does not already exist. The
file is mapped by the runtime linker. In the above examples, this data file is
/var/tmp/libc.so.1.profile. 64bit libraries require an extended profile format
and are written using the .profilex suffix. You can also specify an alternative
directory to store the profile data using the LD_PROFILE_OUTPUT environment
variable.
This profile data file is used to deposit profil(2) data and call count information
related to the use of the specified shared object. This profiled data can be directly
examined with gprof(1).
Note gprof(1) is most commonly used to analyze the gmon.out profile data created
by an executable that has been compiled with the -xpg option of cc(1). The runtime
linkers profile analysis does not require any code to be compiled with this option.
Applications whose dependent shared objects are being profiled should not make calls
to profil(2), because this system call does not provide for multiple invocations
within the same process. For the same reason, these applications must not be compiled
with the -xpg option of cc(1). This compiler-generated mechanism of profiling is also
built on top of profil(2).
One of the most powerful features of this profiling mechanism is to enable the analysis
of a shared object as used by multiple applications. Frequently, profiling analysis is
carried out using one or two applications. However, a shared object, by its very
nature, can be used by a multitude of applications. Analyzing how these applications
use the shared object can offer insights into where energy might be spent to
improvement the overall performance of the shared object.
called/total parents
index %time self descendents called+self name index
called/total children
.....
-----------------------------------------------
0.33 0.00 52/29381 _gettxt [96]
1.12 0.00 174/29381 _tzload [54]
The special name <external> indicates a reference from outside of the address range of
the shared object being profiled. Thus, in the above example, 1634 calls to the function
open(2) within libc occurred from the dynamic executables, or from other shared
objects, bound with libc while the profiling analysis was in progress.
Note The profiling of shared objects is multithread safe, except in the case where one
thread calls fork(2) while another thread is updating the profile data information.
The use of fork1(2) removes this restriction.
ELF objects processed by the link-editors provide many global symbols to which other
objects can bind. These symbols describe the objects application binary interface
(ABI). During the evolution of an object, this interface can change due to the addition
or deletion of global symbols. In addition, the objects evolution can involve internal
implementation changes.
This chapter describes how an objects ABI can be defined and classifies how changes
to this interface can affect backward compatibility. It also presents models by which
interface and implementation changes can be incorporated into new releases of the
object.
The focus of this chapter is on the runtime interfaces of dynamic executables and
shared objects. The techniques used to describe and manage changes within these
dynamic objects are presented in generic terms. A common set of naming conventions
and versioning scenarios as applied to shared objects can be found in Appendix B.
The global symbols made available by any dynamic object represent the objects public
interface. Frequently, the number of global symbols remaining in an object at the end
of a link-edit are more than you would like to make public. These global symbols
result from the relationship required between relocatable objects used to create the
object. They represent private interfaces within the object itself.
Before defining an objects binary interface, you should first determine those global
symbols you wish to make publicly available from the object being created. These
public symbols can be established using the link-editors -M option and an associated
mapfile as part of the final link-edit. This technique is introduced in Reducing
129
Symbol Scope on page 52. This public interface establishes one or more version
definitions within the object being created. These definitions form the foundation for
the addition of new interfaces as the object evolves.
The following sections build upon this initial public interface. First though, you
should understand how various changes to an interface can be categorized so that
they can be managed appropriately.
Interface Compatibility
Many types of change can be made to an object. In their simplest terms, these changes
can be categorized into one of two groups:
Compatible updates. These updates are additive, in that all previously available
interfaces remain intact.
Incompatible updates. These updates have changed the existing interface in such a
way that existing users of the interface can fail or behave incorrectly.
Internal Versioning
A dynamic object can have one or more internal version definitions associated with it.
Each version definition is commonly associated with one or more symbol names. A
symbol name can only be associated with one version definition. However, a version
definition can inherit the symbols from other version definitions. Thus, a structure
exists to define one or more independent, or related, version definitions within the
object being created. As new changes are made to the object, new version definitions
can be added to express these changes.
There are two consequences of providing version definitions within a shared object:
Dynamic objects that are built against a versioned shared object can record their
dependency on the version definitions bound to. These version dependencies are
verified at runtime to ensure that the appropriate interfaces, or functionality, are
available for the correct execution of an application.
Dynamic objects can select the version definitions of a shared object to bind to
during their link-edit. This mechanism enables developers to control their
dependency on a shared object to the interfaces, or functionality, that provide the
most flexibility.
void foo1()
$ cat data.c
const char * _foo1 = "string used by foo1()\n";
$ cat mapfile
SUNW_1.1 { # Release X
global:
foo1;
local:
*;
};
$ cc -o libfoo.so.1 -M mapfile -G foo.o data.o
$ nm -x libfoo.so.1 | grep "foo.$"
[33] |0x0001058c|0x00000004|OBJT |LOCL |0x0 |17 |_foo1
[35] |0x00000454|0x00000034|FUNC |GLOB |0x0 |9 |foo1
The symbol foo1 is the only global symbol defined to provide the shared objects
public interface. The special auto-reduction directive * causes the reduction of all
other global symbols to have local binding within the object being generated. This
directive is introduced in Defining Additional Symbols on page 46. The associated
version name, SUNW_1.1, causes the generation of a version definition. Thus, the
shared objects public interface consists of the internal version definition SUNW_1.1,
associated with the global symbol foo1.
The version definitions contained within an object can be displayed using pvs(1) with
the -d option:
$ pvs -d libfoo.so.1
libfoo.so.1;
SUNW_1.1;
Starting with this initial version definition, the object can evolve by adding new
interfaces and updated functionality. For example, a new function, foo2, together
with its supporting data structures, can be added to the object by updating the source
files foo.c and data.c:
void foo1()
{
(void) printf(_foo1);
}
void foo2()
{
(void) printf(_foo2);
}
$ cat data.c
const char * _foo1 = "string used by foo1()\n";
const char * _foo2 = "string used by foo2()\n";
The creation of this new interface is important as it identifies the evolution of the
object and enables users to verify and select the interfaces to which they bind. These
concepts are covered in more detail in Binding to a Version Definition on page 136
and in Specifying a Version Binding on page 140.
The following example shows the mapfile directives that create these two interfaces.
$ cat mapfile
SUNW_1.1 { # Release X
global:
foo1;
local:
*;
};
The symbols foo1 and foo2 are both defined to be part of the shared objects public
interface. However, each of these symbols is assigned to a different version definition;
foo1 is assigned to SUNW_1.1, and foo2 is assigned to SUNW_1.2.
The inheritance of one version definition by another is a useful technique that reduces
the version information that will eventually be recorded by any object that binds to a
version dependency. Version inheritance is covered in more detail in the section
Binding to a Version Definition on page 136.
Any internal version definition has an associated version definition symbol created. As
shown in the previous pvs(1) example, these symbols are displayed when using the
-v option.
Such a version definition is empty, in that it has no global interface symbols associated
with it.
For example, suppose the data file data.c, used in the previous examples, is updated
to provide more detailed string definitions:
$ cat data.c
const char * _foo1 = "string used by function foo1()\n";
const char * _foo2 = "string used by function foo2()\n";
The empty version definition is signified by the weak label. These weak version
definitions enable applications to verify the existence of a particular implementation
by binding to the version definition associated with that functionality. The section
Binding to a Version Definition on page 136 illustrates how these definitions can be
used in more detail.
void bar1()
{
foo1();
}
$ cat bar2.c
extern void foo2();
void bar2()
{
foo2();
}
These two symbols are intended to define two new public interfaces. Neither of these
new interfaces are related to each other. However, each expresses a dependency on the
original SUNW_1.2 interface.
Again, the version definitions created in libfoo.so.1 when using this mapfile,
and their related dependencies, can be inspected using pvs(1):
$ cc -o libfoo.so.1 -M mapfile -G foo.o bar1.o bar2.o data.o
$ pvs -dv libfoo.so.1
libfoo.so.1;
SUNW_1.1;
SUNW_1.2: {SUNW_1.1};
SUNW_1.2.1 [WEAK]: {SUNW_1.2};
SUNW_1.3a: {SUNW_1.2};
SUNW_1.3b: {SUNW_1.2};
The following sections explore how these version definition recordings can be used to
verify runtime binding requirements and control the binding of an object during its
creation.
The following example takes the data files from the previous section and generates a
shared object suitable for a compile time environment. This shared object,
libfoo.so.1, is used in the succeeding binding examples.
In effect, there are six public interfaces being offered by the shared object. Four of these
interfaces, SUNW_1.1, SUNW_1.2, SUNW_1.3a, and SUNW_1.3b, define exported
symbol names. One interface, SUNW_1.2.1, describes an internal implementation
change to the shared object, and one interface, libfoo.so.1, defines several reserved
labels. Dynamic objects created with this shared object as a dependency, record the
version names of the interfaces the dynamic object binds to.
The following example creates an application that references symbols foo1 and foo2.
The versioning dependency information recorded in the application can be examined
using pvs(1) with the -r option.
$ cat prog.c
extern void foo1();
extern void foo2();
main()
{
foo1();
foo2();
}
$ cc -o prog prog.c -L. -R. -lfoo
$ pvs -r prog
libfoo.so.1 (SUNW_1.2, SUNW_1.2.1);
In this example, the application prog has bound to the two interfaces SUNW_1.1 and
SUNW_1.2. These interfaces provided the global symbols foo1 and foo2 respectively.
Because the application prog was built against the shared objects implementation
containing the weak version definition SUNW_1.2.1, this dependency is also
recorded. Even though this version definition is defined to inherit the version
definition SUNW_1.2, the versions weak nature precludes its normalization with
SUNW_1.1, and results in a separate dependency recording.
Had there been multiple weak version definitions that inherited from each other, then
these definitions will be normalized in the same manner as non-weak version
definitions are.
Having recorded these version definition dependencies, the runtime linker validates
the existence of the required version definitions in the objects that are bound to when
the application is executed. This validation can be displayed using ldd(1) with the -v
option. For example, by running ldd(1) on the application prog, the version
definition dependencies are shown to be found correctly in the shared object
libfoo.so.1:
$ ldd -v prog
Note ldd(1) with the -v option implies verbose output. A recursive list of all
dependencies, together with all versioning requirements, is generated.
Had the application prog not recorded any version definition dependencies, the
nonexistence of the required interface symbol foo2 would have manifested itself
some time during the execution of the application as a fatal relocation error. This
relocation error might occur at process initialization, during process execution, or
might not occur at all if the execution path of the application did not call the function
foo2. See Relocation Errors on page 72.
main()
{
void * handle;
const char * file = "libfoo.so.1";
const char * vers = "SUNW_1.2";
....
An objects binding requirements can be controlled using a file control directive. This
directive is supplied using the link-editors -M option and an associated mapfile. The
following syntax for file control directives is available:
name - version [ version ... ] [ $ADDVERS=version ];
name Represents the name of the shared object dependency. This name should
match the shared objects compilation environment name as used by the
link-editor. See Library Naming Conventions on page 31.
version Represents the version definition name within the shared object that
should be made available for binding. Multiple version definitions can be specified.
$ADDVERS Allows for additional version definitions to be recorded.
The following example illustrates the use of the version control mechanism. This
example uses the shared object libfoo.so.1 containing the following version
interface definitions:
$ pvs -dsv libfoo.so.1
libfoo.so.1:
_end;
_GLOBAL_OFFSET_TABLE_;
_DYNAMIC;
_edata;
_PROCEDURE_LINKAGE_TABLE_;
_etext;
SUNW_1.1:
foo1;
foo2;
SUNW_1.1;
SUNW_1.2: {SUNW_1.1}:
bar;
For example, suppose you develop an application, prog, and want to ensure that the
application can run on Release X. The application can then only use the interfaces
available in that release. If the application mistakenly references the symbol bar, then
the application is not compliant with the required interface. This condition is signalled
by the link-editor as an undefined symbol error:
$ cat prog.c
extern void foo1();
extern void bar();
main()
{
foo1();
bar();
}
To be compliant with the SUNW_1.1 interface, you must remove the reference to bar.
You can either rework the application to remove the requirement on bar, or add an
implementation of bar to the creation of the application.
From the previous libfoo.so.1 example, assume that in Release X+2, the
version definition SUNW_1.1 is subdivided into two standard releases, STAND_A and
STAND_B. To preserve compatibility, the SUNW_1.1 version definition must be
maintained. In this example, this version definition is expressed as inheriting the two
standard definitions:
$ pvs -dsv libfoo.so.1
libfoo.so.1:
_end;
_GLOBAL_OFFSET_TABLE_;
_DYNAMIC;
_edata;
_PROCEDURE_LINKAGE_TABLE_;
_etext;
SUNW_1.1: {STAND_A, STAND_B}:
SUNW_1.1;
SUNW_1.2: {SUNW_1.1}:
bar;
STAND_A:
foo1;
STAND_A;
STAND_B:
foo2;
STAND_B;
The application prog can be built to align its requirement with previous releases by
creating a dependency on SUNW_1.1:
$ cat mapfile
libfoo.so - SUNW_1.1 $ADDVERS=SUNW_1.1;
$ cat prog
extern void foo1();
main()
{
foo1();
}
$ cc -M mapfile -o prog prog.c -L. -R. -lfoo
$ pvs -r prog
libfoo.so.1 (SUNW_1.1);
Creating a Weak Version Definition on page 134 described how weak version
definitions can be used to mark an internal implementation change. These version
definitions are well suited to indicate bug fixes and performance improvements made
to an object. If the existence of a weak version is required, an explicit dependency on
this version definition can be generated. The creation of such a dependency can be
important when a bug fix, or performance improvement, is critical for the object to
function correctly.
From the previous libfoo.so.1 example, assume a bug fix is incorporated as the
weak version definition SUNW_1.2.1 in software Release X+3:
$ pvs -dsv libfoo.so.1
libfoo.so.1:
_end;
_GLOBAL_OFFSET_TABLE_;
_DYNAMIC;
_edata;
_PROCEDURE_LINKAGE_TABLE_;
_etext;
SUNW_1.1: {STAND_A, STAND_B}:
SUNW_1.1;
SUNW_1.2: {SUNW_1.1}:
bar;
STAND_A:
foo1;
STAND_A;
STAND_B:
foo2;
Normally, if an application is built against this shared object, the application records a
weak dependency on the version definition SUNW_1.2.1. This dependency is
informational only. This dependency does not cause termination of the application
should the version definition not exist in the libfoo.so.1 used at runtime.
The file control directive $ADDVERS can be used to generate an explicit dependency on
a version definition. If this definition is weak, then this explicit reference also causes
the version definition to be promoted to a strong dependency.
The application prog can be built to enforce the requirement that the SUNW_1.2.1
interface be available at runtime by using the following file control directive:
$ cat mapfile
libfoo.so - SUNW_1.1 $ADDVERS=SUNW_1.2.1;
$ cat prog
extern void foo1();
main()
{
foo1();
}
$ cc -M mapfile -o prog prog.c -L. -R. -lfoo
$ pvs -r prog
libfoo.so.1 (SUNW_1.2.1);
prog has been built with an explicit dependency on the interface STAND_A. Because
the version definition SUNW_1.2.1 is promoted to a strong version, it is also
normalized with the dependency STAND_A. At runtime, if the version definition
SUNW_1.2.1 cannot be found, a fatal error is generated.
Note When working with a small number of dependencies, you can use the
link-editors -u option to explicitly bind to a version definition. Use this option to
reference the version definition symbol. However, a symbol reference is nonselective.
When working with multiple dependencies, that might contain similarly named
version definitions, this technique may be insufficient to create explicit bindings.
Version Stability
The various models for binding to versions within an object only remain intact if the
individual version definitions remain constant over the life time of the object.
Relocatable Objects
Version information can be recorded and used within dynamic objects. Relocatable
objects can maintain versioning information in a similar manner. However, there are
some subtle differences in how this information is used.
Any version definitions supplied to the link-edit of a relocatable object are recorded in
the same format as they are when building dynamic executables or shared objects.
However, by default, symbol reduction is not carried out on the object being created.
Instead, when the relocatable object is finally used as input to the generation of a
dynamic object, the version recording itself will be used to determine the symbol
reductions to apply.
In addition, any version definitions found in relocatable objects are propagated to the
dynamic object. For an example of version processing in relocatable objects, see
Reducing Symbol Scope on page 52.
External Versioning
Runtime references to a shared object should always refer to the files version file
name. This is usually expressed as a file name with a version number suffix. When a
shared objects interface changes in an incompatible manner, such that it will break old
applications, a new shared object should be distributed with a new versioned file
name. In addition, the original versioned file name must still be distributed to provide
the interfaces required by the old applications.
You should provide shared objects as separate versioned file names within the runtime
environment when building applications over a series of software releases. You can
then guarantee that the interface against which the applications were built is available
for them to bind during their execution.
The following section describes how to coordinate the binding of an interface between
the compilation and runtime environments.
However, at runtime, any shared object dependencies should exist in their versioned
name form. Instead of maintaining two distinct shared objects that follow these
naming conventions, create file system links between the two file names.
Either a symbolic link or hard link can be used. However, as a documentation and
diagnostic aid, symbolic links are more useful.
The shared object libfoo.so.1 has been generated for the runtime environment.
Generating a symbolic link libfoo.so, has also enabled this files use in a
compilation environment. For example:
$ cc -o prog main.o -L. -lfoo
The link-editor processes the relocatable object main.o with the interface described by
the shared object libfoo.so.1, which is found by following the symbolic link
libfoo.so.
Over a series of software releases, new versions of this shared object may be
distributed with changed interfaces. The compilation environment can be constructed
to use the interface that is applicable by changing the symbolic link. For example:
$ ls -l libfoo*
lrwxrwxrwx 1 usr grp 11 1993 libfoo.so -> libfoo.so.3
-rwxrwxr-x 1 usr grp 3136 1991 libfoo.so.1
-rwxrwxr-x 1 usr grp 3237 1992 libfoo.so.2
-rwxrwxr-x 1 usr grp 3554 1993 libfoo.so.3
Three major versions of the shared object are available. Two of these shared objects,
libfoo.so.1 and libfoo.so.2, provide the dependencies for existing
applications. libfoo.so.3 offers the latest major release for creating and running
new applications.
prog:
**** DYNAMIC SECTION INFORMATION ****
.dynamic:
[INDEX] Tag Value
[1] NEEDED libfoo.so
.........
When the application prog is executed, the runtime linker searches for the
dependency libfoo.so. prog binds to the file to which this symbolic link is
pointing.
To provide the correct runtime name to be recorded as a dependency, the shared object
libfoo.so.1 should be built with an soname definition. This definition identifies the
shared objects runtime name. This name is used as the dependency name by any
object that links against this shared object. This definition can be provided using the
-h option during the link-edit of the shared object itself. For example:
$ cc -o libfoo.so.1 -G -K pic -h libfoo.so.1 foo.c
$ ln -s libfoo.so.1 libfoo.so
$ cc -o prog main.o -L. -lfoo
$ dump -Lv prog
prog:
**** DYNAMIC SECTION INFORMATION ****
.dynamic:
[INDEX] Tag Value
[1] NEEDED libfoo.so.1
.........
This symbolic link and the soname mechanism have established a robust coordination
between the shared-object naming conventions of the compilation and runtime
environment. The interface processed during the link-edit is accurately recorded in the
output file generated. This recording ensures that the intended interface are furnished
at runtime.
Support Interfaces
The link-editors provide a number of support interfaces that enable the monitoring,
and in some cases modification, of link-editor and runtime linker processing. These
interfaces typically require a more advanced understanding of link-editing concepts
than has been described in previous chapters. The following interfaces are described
in this chapter:
ld-support Link-Editor Support Interface on page 149
rtld-audit Runtime Linker Auditing Interface on page 155
rtld-debugger Runtime Linker Debugger Interface on page 164
This section describes the ld-support interface for input file inspection, and to some
degree, input file data modification of those files that compose a link-edit. Two
applications that employ this interface are the link-editor itself, which uses it to
process debugging information within relocatable objects, and the make(1S) utility,
which uses it to save state information.
The ld-support interface is composed of a support library that offers one or more
support interface routines. This library is loaded as part of the link-edit process, and
any support routines found are called at various stages of link-editing.
You should be familiar with the elf(3ELF) structures and file format when using this
interface.
149
Invoking the Support Interface
The link-editor accepts one or more support libraries provided by either the
SGS_SUPPORT environment variable or with the link-editors -S option. The
environment variable consists of a colon separated list of support libraries:
$ SGS_SUPPORT=./support.so.1:libldstab.so.1 cc ...
The -S option specifies a single support library. Multiple -S options can be specified:
$ LD_OPTIONS="-S./support.so.1 -Slibldstab.so.1" cc ...
A support library is a shared object. The link-editor opens each support library, in the
order they are specified, using dlopen(3DL). If both the environment variable and -S
option are encountered, then the support libraries specified with the environment
variable are processed first. Each support library is then searched, using dlsym(3DL),
for any support interface routines. These support routines are then called at various
stages of link-editing.
A support library must be consistent with the ELF class of the link-editor being
invoked, either 32bit or 64bit. See 32Bit and 64Bit Environments on page 150 for
more details.
The support interface for 64bit objects is similar to that of 32bit objects, but ends in a
64 suffix, for example ld_start() and ld_start64(). This convention allows both
implementations of the support interface to reside in a single shared object
libldstab.so.1 of each class, 32bit and 64bit.
The SGS_SUPPORT environment variable can be specified with a _32 or _64 suffix,
and the link-editor options -z ld32 and -z ld64 can be used to define -S option
requirements. These definitions will only be interpreted, respectively, by the 32bit or
64bit class of the link-editor. This enables both classes of support library to be
specified when the class of the link-editor may not be known.
The link-editor calls this interface with the highest version of the ld-support
interface it is capable of supporting. The support library can verify that this version
is sufficient for its use, and return the version it expects to use. This version is
normally LD_SUP_VCURRENT.
If the support library does not provide this interface, the initial support level
LD_SUP_VERSION1 is assumed.
If the support library returns a version of zero, or a value greater than the
ld-support interface the link-editor supports, the support library will not be used.
ld_start()
This function is called after initial validation of the link-editor command line, and
indicates the start of input file processing.
void ld_start(const char * name, const Elf32_Half type,
const char * caller);
name is the output file name being created. type is the output file type, which is
either ET_DYN, ET_REL, or ET_EXEC, as defined in sys/elf.h. caller is the
application calling the interface, which is normally /usr/ccs/bin/ld.
ld_file()
This function is called for each input file before any processing of the files data is
carried out.
void ld_file(const char * name, const Elf_Kind kind, int flags,
Elf * elf);
name is the input file about to be processed. kind indicates the input file type, which
is either ELF_K_AR, or ELF_K_ELF, as defined in libelf.h. flags indicates how
the link-editor obtained the file, and can be one or more of the following definitions:
If no flags values are specified then the input file has been explicitly named on the
command line. elf is a pointer to the files ELF descriptor.
ld_input_section()
This function is called for each section of the input file. This function is called
before the link-editor has determined whether the section should be propagated to
the output file. This function differs from ld_section() processing, which is only
called for sections that contribute to the output file.
void ld_input_section(const char * name, Elf32_Shdr ** shdr,
Elf32_Word sndx, Elf_Data * data, Elf * elf, unit_t flags);
name is the input section name. shdr is a pointer to the associated section header.
sndx is the section index within the input file. data is a pointer to the associated data
buffer. elf is a pointer to the files ELF descriptor. flags is reserved for future use.
You can modify the data by reallocating the data and reassigning the Elf_Data
buffers d_buf pointer. Any modification to the data should ensure the correct
setting of the Elf_Data buffers d_size element. For input sections that become
part of the output image, setting the d_size element to zero effectively removes
the data from the output image.
The flags field points to a uint_t data field that is initially zero filled. No flags are
currently assigned, although the ability to assign flags in future updates, by the
link-editor or the support library, is provided.
ld_section()
This function is called for each section of the input file that will be propagated to
the output file, but before any processing of the section data is carried out.
void ld_section(const char * name, Elf32_Shdr * shdr,
Elf32_Word sndx, Elf_Data * data, Elf * elf);
You can modify the data by reallocating the data itself and reassigning the
Elf_Data buffers d_buf pointer. Any modification to the data should ensure the
correct setting of the Elf_Data buffers d_size element. For input sections that
will become part of the output image, setting the d_size element to zero will
effectively remove the data from the output image.
Note Sections that are stripped using the link-editors -s option, or discarded due
to SHT_SUNW_COMDAT processing or SHF_EXCLUDE identification, are not reported
to ld_section(). See COMDAT Section on page 203, and Table 714.
ld_input_done()
This function is called when input file processing is complete but before the output
file is laid out.
void ld_input_done(uint_t flags);
The flags field points to a uint_t data field that is initially zero filled. No flags are
currently assigned, although the ability to assign flags in future updates, by the
link-editor or the support library, is provided.
ld_atexit()
This function is called when the link-edit is complete.
void ld_atexit(int status);
status is the exit(2) code that will be returned by the link-editor and is either
EXIT_FAILURE or EXIT_SUCCESS, as defined in stdlib.h.
void
ld_start(const char * name, const Elf32_Half type,
const char * caller)
void
ld_file(const char * name, const Elf_Kind kind, int flags,
Elf * elf)
{
if (flags & LD_SUP_EXTRACTED)
indent = 4;
else
indent = 2;
void
ld_section(const char * name, Elf32_Shdr * shdr, Elf32_Word sndx,
Elf_Data * data, Elf * elf)
{
Elf32_Ehdr * ehdr = elf32_getehdr(elf);
if (ehdr->e_type == ET_REL)
(void) printf("%*s section [%ld]: %s\n", indent,
"", (long)sndx, name);
}
This support library is dependent upon libelf to provide the ELF access function
elf32_getehdr(3ELF) that is used to determine the input file type. The support
library is built using:
$ cc -o support.so.1 -G -K pic support.c -lelf -lc
The following example shows the section diagnostics resulting from the construction
of a trivial application from a relocatable object and a local archive library. The
invocation of the support library, in addition to default debugging information
processing, is brought about by the -S option usage.
$ LD_OPTIONS="-S./support.so.1 -Slibldstab.so.1" \
cc -o prog main.c -L. -lfoo
Note The number of sections displayed in this example have been reduced to
simplify the output. Also, the files included by the compiler driver can vary.
The rtld-audit interface is implemented as an audit library that offers one or more
auditing interface routines. If this library is loaded as part of a process, the audit
routines are called by the runtime linker at various stages of process execution. These
interfaces enable the audit library to access:
The search for dependencies. Search paths may be substituted by the audit library.
Information regarding loaded objects.
Symbol bindings that occur between loaded objects. These bindings can be altered
by the audit library.
Exploitation of the lazy binding mechanism provided by procedure linkage table
entries to allow auditing of function calls and their return values. The arguments to
a function and its return value can be modified by the audit library. See Procedure
Linkage Table (Processor-Specific) on page 261.
The runtime linker itself is also described by a link-map. This link-map is maintained
on a different list from that of the application objects. The runtime linker therefore
resides in its own unique name space, which prevents any direct binding of the
application to services within the runtime linker. An application can only call upon the
public services of the runtime linker by the filter libdl.so.1.
The rtld-audit interface employs its own link-map list on which it maintains any audit
libraries. The audit libraries are thus isolated from the symbol binding requirements of
the application. Inspection of the application link-map list is possible with
dlmopen(3DL). When used with the RTLD_NOLOAD flag, dlmopen(3DL) allows the
audit library to query an objects existence without causing its loading.
If the audit library calls printf(3C), then the audit library must define a dependency
on libc. See Generating a Shared Object Output File on page 44. Because the audit
library has a unique namespace, symbol references cannot be satisfied by the libc
present in the application being audited. If an audit library has a dependency on
libc, then two versions of libc.so.1 are loaded into the process. One version
satisfies the binding requirements of the application link-map list. The other version
satisfies the binding requirements of the audit link-map list.
To ensure that audit libraries are built with all dependencies recorded, use the
link-editors -z defs option.
The rtld-audit interface enables multiple audit libraries to be supplied. Audit libraries
that expect to be employed in this fashion should not alter the bindings that would
normally be returned by the runtime linker. Alteration of these bindings can produce
unexpected results from audit libraries that follow.
Secure applications can only obtain audit libraries from trusted directories. By default,
the only trusted directory known to the runtime linker for 32bit objects is
/usr/lib/secure. For 64bit objects, the trusted directory is
/usr/lib/secure/64.
With this mechanism alone, information such as searching for the identifying object
has occurred prior to the audit library being loaded. To provide as much auditing
information as possible, the existence of an object requiring local auditing is
propagated to users of that object. For example, if an application is built that depends
on libfoo.so.1, then the application is identified to indicate its dependencies
require auditing:
$ cc -o main main.c libfoo.so.1
$ dump -Lv main | fgrep AUDIT
[5] DEPAUDIT audit.so.1
The auditing enabled via this mechanism results in the audit library being passed
information regarding all of the applications explicit dependencies. This dependency
auditing can also be recorded directly when creating an object by using the
link-editors -P option:
$ cc -o main main.c -Wl,-Paudit.so.1
$ dump -Lv main | fgrep AUDIT
[5] DEPAUDIT audit.so.1
Note References to architecture, or object class specific interfaces are reduced to their
generic name to simplify the discussions. For example, a reference to
la_symbind32() and la_symbind64() is specified as la_symbind().
la_version()
This function provides the initial handshake between the runtime linker and the
audit library. This interface must be provided for the audit library to be loaded.
uint_t la_version(uint_t version);
The runtime linker calls this interface with the highest version of the rtld-audit
interface it is capable of supporting. The audit library can verify that this version is
sufficient for its use, and return the version it expects to use. This version is
normally LAV_CURRENT, which is defined in /usr/include/link.h.
cookie identifies the object heading the link-map. flags indicates the type of activity
as defined in /usr/include/link.h:
LA_ACT_ADD Objects are being added to the link-map list.
LA_ACT_DELETE Objects are being deleted from the link-map list.
LA_ACT_CONSISTENT Object activity has been completed.
la_objsearch()
This function informs an auditor that an object is about to be searched for.
char * la_objsearch(const char * name, uintptr_t * cookie, uint_t flags);
name indicates the file or path name being searched for. cookie identifies the object
initiating the search. flags identifies the origin and creation of name as defined in
/usr/include/link.h:
LA_SER_ORIG This is the initial search name. Typically, this indicates the file
name that is recorded as a DT_NEEDED entry, or the argument supplied to
dlmopen(3DL).
LA_SER_LIBPATH The path name has been created from a
LD_LIBRARY_PATH component.
LA_SER_RUNPATH The path name has been created from a runpath
component.
LA_SER_DEFAULT The path name has been created from a default search path
component.
LA_SER_CONFIG The path component originated from a configuration file.
See the crle(1) man page.
LA_SER_SECURE The path component is specific to secure objects.
The return value indicates the search path name that the runtime linker should
continue to process. A value of zero indicates that this path should be ignored. An
audit library that simply monitors search paths should return name.
la_objopen()
This function is called each time a new object is loaded by the runtime linker.
uint_t la_objopen(Link_map * lmp, Lmid_t lmid, uintptr_t * cookie);
lmp provides the link-map structure that describes the new object. lmid identifies the
link-map list to which the object has been added. cookie provides a pointer to an
identifier. This identifier is initialized to the objects lmp. This identifier can be
modified by the audit library to better identify the object to other rtld-audit
interface routines
See the la_symbind() function for more details on the use of these two flags.
A return value of zero indicates that binding information is of no interest for this
object.
la_preinit()
This function is called once after all objects have been loaded for the application,
but before transfer of control to the application occurs.
void la_preinit(uintptr_t * cookie);
cookie identifies the primary object that started the process, normally the dynamic
executable.
la_symbind()
This function is called when a binding occurs between two objects that have been
tagged for binding notification from la_objopen().
uintptr_t la_symbind32(Elf32_Sym * sym, uint_t ndx,
uintptr_t * refcook, uintptr_t * defcook, uint_t * flags);
ndx indicates the symbol index within the bound objects dynamic symbol table.
refcook describes the object making reference to this symbol. This identifier is the
same identifier as passed to the la_objopen() function that returned
LA_FLG_BINDFROM. defcook describes the object defining this symbol. This
identifier is the same as passed to the la_objopen() that returned
LA_FLG_BINDTO.
flags points to a data item that can convey information regarding the binding. This
data item can be used to modify the continued auditing of procedure linkage table
entries. This value is a mask of the following flags defined in
/usr/include/link.h:
LA_SYMB_NOPLTENTER The la_pltenter() function is not called for this
symbol.
The return value indicates the address to which control should be passed following
this call. An audit library that simply monitors symbol binding should return the
value of sym->st_value so that control is passed to the bound symbol definition.
An audit library can intentionally redirect a symbol binding by returning a different
value.
sym_name, which is applicable for la_symbind64() only, contains the name of the
symbol being processed. This name is available in the sym->st_name field for the
32bit interface.
la_pltenter()
These functions are called on a SPARC and x86 system respectively. These functions
are called when a procedure linkage table entry, between two objects that have been
tagged for binding notification, is called.
uintptr_t la_sparcv8_pltenter(Elf32_Sym * sym, uint_t ndx,
uintptr_t * refcook, uintptr_t * defcook,
La_sparcv8_regs * regs, uint_t * flags);
sym, ndx, refcook, defcook and sym_name provide the same information as passed to
la_symbind().
regs points to the out registers on a SPARC system, and the stack and frame
registers on a x86 system, as defined in /usr/include/link.h.
flags points to a data item that can convey information regarding the binding. This
data can be used to modify the continuing auditing of this procedure linkage table
entry. This data item is the same as pointed to by the flags from la_symbind().
This value is a mask of the following flags defined in /usr/include/link.h:
LA_SYMB_NOPLTENTER la_pltenter() is not be called again for this
symbol.
The return value indicates the address to which control should be passed following
this call. An audit library that simply monitors symbol binding should return the
value of sym->st_value so that control is passed to the bound symbol definition.
An audit library can intentionally redirect a symbol binding by returning a different
value.
la_pltexit()
This function is called when a procedure linkage table entry, between two objects
that have been tagged for binding notification, returns. This function is called
before control reaches the caller.
uintptr_t la_pltexit(Elf32_Sym * sym, uint_t ndx, uintptr_t * refcook,
uintptr_t * defcook, uintptr_t retval);
sym, ndx, refcook, defcook and sym_name provide the same information as passed to
la_symbind(). retval is the return code from the bound function. An audit library
that simply monitors symbol binding should return retval. An audit library can
intentionally return a different value.
la_objclose()
This function is called after any termination code for an object has been executed
and prior to the object being unloaded.
uint_t la_objclose(uintptr_t * cookie);
cookie identifies the object, and was obtained from a previous la_objopen(). Any
return value is presently ignored.
uint_t
la_version(uint_t version)
{
uint_t
la_objopen(Link_map * lmp, Lmid_t lmid, uintptr_t * cookie)
{
if (lmid == LM_ID_BASE)
(void) printf("file: %s loaded\n", lmp->l_name);
return (0);
}
$ cc -o audit.so.1 -G -K pic -z defs audit.c -lmapmalloc -lc
$ LD_AUDIT=./audit.so.1 date
file: date loaded
file: /usr/lib/libc.so.1 loaded
file: /usr/lib/libdl.so.1 loaded
file: /usr/lib/locale/en_US/en_US.so.2 loaded
Thur Aug 10 17:03:55 PST 2000
sotruss(1) and whocalls(1) are included in the SUNWtoo package. perfcnt and
symbindrep are example programs. They are not intended for use in a production
environment.
The runtime linker cannot detect functions of this type, and thus the audit library
creator is responsible for disabling la_pltexit() for such routines.
This section describes the rtld-debugger interface for monitoring and modifying a
dynamically linked application from another process. The architecture of this interface
follows the model used in libthread_db(3THR).
When using the rtld-debugger interface, at least two processes are involved:
One or more target processes. The target processes must be dynamically linked and
use the runtime linker /usr/lib/ld.so.1 for 32bit processes, or
/usr/lib/64/ld.so.1 for 64bit processes.
A controlling process links with the rtld-debugger interface library and uses it to
inspect the dynamic aspects of the target processes. A 64bit controlling process
can debug both 64bit and 32bit targets. However, a 32bit controlling process is
limited to 32bit targets.
The most anticipated use of the rtld-debugger interface is when the controlling process
is a debugger and its target is a dynamic executable.
The rtld-debugger interface enables the following activities with a target process:
Initial rendezvous with the runtime linker.
Notification of the loading and unloading of dynamic objects.
Retrieval of information regarding any loaded objects.
Stepping over procedure linkage table entries.
The rtld-debugger interface assumes that the process being analyzed is stopped when
requests are made of the rtld-debugger interface. If this halt does not occur, data
structures within the runtime linker of the target process might not be in a consistent
state for examination.
Dynamic
Debugger application
Process data
/proc
php is a cookie created by the controlling process to identify the target process. This
cookie is used by the imported interface offered by the controlling process to
maintain context, and is opaque to the rtld-debugger interface.
rd_reset()
This function resets the information within the agent based off the same
ps_prochandle structure given to rd_new().
rd_err_e rd_reset(struct rd_agent * rdap);
Error Handling
The following error states can be returned by the rtld-debugger interface (defined in
rtld_db.h):
typedef enum {
RD_ERR,
RD_OK,
RD_NOCAPAB,
RD_DBERR,
RD_NOBASE,
RD_NODYNAM,
RD_NOMAPS
} rd_err_e;
rd_log()
This function turns logging on (1) or off (0).
void rd_log(const int onoff);
When logging is turned on, the imported interface function ps_plog() provided
by the controlling process, is called with more detailed diagnostic information.
Notice that all addresses given in this structure, including string pointers, are
addresses in the target process and not in the address space of the controlling process
itself.
rl_nameaddr
A pointer to a string that contains the name of the dynamic object.
rl_flags
With revision RD_VERSION2, dynamically loaded relocatable objects are identified
with RD_FLG_MEM_OBJECT.
rl_base
The base address of the dynamic object.
rl_data_base
The base address of the data segment of the dynamic object.
rl_lmident
The link-map identifier (see Establishing a Namespace on page 156).
rl_refnameaddr
If the dynamic object is a standard filter, then this points to the name of the filtees.
rl_plt_base, rl_plt_size
These elements are present for backward compatibility and are currently unused.
rl_bend
The end address of the object (text + data + bss). With revision RD_VERSION2,
a dynamically loaded relocatable object will cause this element to point to the end
of the created object, which will include its section headers.
rl_padstart
The base address of the padding before the dynamic object (refer to Dynamic
Object Padding on page 174).
rl_padend
The base address of the padding after the dynamic object (refer to Dynamic Object
Padding on page 174).
rl_dynamic
This field, added with RD_VERSION2, provides the base address of the objects
dynamic section, which allows reference to such entries as DT_CHECKSUM (see Table
742).
Return codes from the cb routine are examined by rd_loadobj_iter() and have
the following meaning:
1 continue processing link-maps.
0 stop processing link-maps and return control to the controlling process.
Event Notification
A controlling process can track certain events that occur within the scope of the
runtime linker that. These events are:
RD_PREINIT
The runtime linker has loaded and relocated all the dynamic objects and is about to
start calling the .init sections of each object loaded.
RD_POSTINIT
The runtime linker has finished calling all of the .init sections and is about to
transfer control to the primary executable.
RD_DLACTIVITY
The runtime linker has been invoked to either load or unload a dynamic object.
These events can be monitored using the following interface, defined in sys/link.h
and rtld_db.h:
typedef enum {
RD_NONE = 0,
RD_PREINIT,
RD_POSTINIT,
RD_DLACTIVITY
} rd_event_e;
/*
* Ways that the event notification can take place:
*/
typedef enum {
RD_NOTIFY_BPT,
RD_NOTIFY_AUTOBPT,
RD_NOTIFY_SYSCALL
} rd_notify_e;
Note Presently, for performance reasons, the runtime linker ignores event
disabling. The controlling process should not assume that a given break-point can
not be reached because of the last call to this routine.
rd_event_addr()
This function specifies how the controlling program is notified of a given event.
rd_err_e rd_event_addr(rd_agent_t * rdap, rd_event_e event,
rd_notify_t * notify);
Depending on the event type, the notification of the controlling process takes place
by calling a benign, cheap system call that is identified by notify-
>u.syscallno, or executing a break point at the address specified by
notify->u.bptaddr. The controlling process is responsible for tracing the system
call or place the actual break-point.
When an event has occurred, additional information can be obtained by this interface,
defined in rtld_db.h:
typedef enum {
RD_NOSTATE = 0,
RD_CONSISTENT,
RD_ADD,
RD_DELETE
} rd_state_e;
The following table shows the possible state for each of the different event types.
RD_ADD
RD_DELETE
The following interface enables a controlling process to step over the runtime linkers
procedure linkage table processing. The controlling process can determine when a
procedure linkage table entry is encountered based on external information provided
in the ELF file.
Once a target process has stepped into a procedure linkage table entry, the process
calls the rd_plt_resolution() interface:
pc represents the first instruction of the procedure linkage table entry. lwpid
provides the lwp identifier and plt_base provides the base address of the procedure
linkage table. These three variables provide information sufficient for various
architectures to process the procedure linkage table.
rpi provides detailed information regarding the procedure linkage table entry as
defined in the following data structure, defined in rtld_db.h:
typedef enum {
RD_RESOLVE_NONE,
RD_RESOLVE_STEP,
RD_RESOLVE_TARGET,
RD_RESOLVE_TARGET_STEP
} rd_skip_e;
The following scenarios are possible from the rd_plt_info_t return values:
The first call through this procedure linkage table must be resolved by the runtime
linker. In this case, the rd_plt_info_t contains:
{RD_RESOLVE_TARGET_STEP, M, <BREAK>, 0, 0}
The controlling process sets a breakpoint at BREAK and continues the target process.
When the breakpoint is reached, the procedure linkage table entry processing has
finished. The controlling process can then step M instructions to the destination
function. Notice that the bound address (pi_baddr) has not been set since this is
the first call through a procedure linkage table entry.
On the Nth time through this procedure linkage table, rd_plt_info_t contains:
{RD_RESOLVE_STEP, M, 0, <BoundAddr>, RD_FLG_PI_PLTBOUND}
The procedure linkage table entry has already been resolved and the controlling
process can step M instructions to the destination function. The address that the
procedure linkage table entry is bound to is <BoundAddr> and the
RD_FLG_PI_PLTBOUND bit has been set in the flags field.
padsize specifies the size of the padding, in bytes, to be preserved both before and
after any objects loaded into memory. This padding is reserved as a memory
mapping using mmap(2) with PROT_NONE permissions and the MAP_NORESERVE
flag. Effectively, the runtime linker reserves areas of the virtual address space of the
target process adjacent to any loaded objects. These areas can later be utilized by the
controlling process.
The following interfaces are currently being used by the rtld-debugger interface:
ps_pauxv()
This function returns a pointer to a copy of the auxv vector.
ps_err_e ps_pauxv(const struct ps_prochandle * ph, auxv_t ** aux);
Because the auxv vector information is copied to an allocated structure, the pointer
remains as long as the ps_prochandle is valid.
ps_pread()
This function reads data from the target process.
ps_err_e ps_pread(const struct ps_prochandle * ph, paddr_t addr,
char * buf, int size);
From address addr in the target process, size bytes are copied to buf.
ps_pwrite()
This function writes data to the target process.
ps_err_e ps_pwrite(const struct ps_prochandle * ph, paddr_t addr,
char * buf, int size);
size bytes from buf are copied into the target process at address addr.
ps_plog()
This function is called with additional diagnostic information from the
rtld-debugger interface.
void ps_plog(const char * fmt, ...);
The controlling process determines where, or if, to log this diagnostic information.
The arguments to ps_plog() follow the printf(3C) format.
ps_pglobal_lookup()
This function searches for the symbol in the target process.
The symbol named name is searched for within the object named obj within the
target process ph. If the symbol is found, the symbol address is stored in sym_addr.
ps_pglobal_sym()
This function searches for the symbol in the target process.
ps_err_e ps_pglobal_sym(const struct ps_prochandle * ph,
const char * obj, const char * name, ps_sym_t * sym_desc);
The symbol named name is searched for within the object named obj within the
target process ph. If the symbol is found, the symbol descriptor is stored in
sym_desc.
In the event that the rtld-debugger interface needs to find symbols within the
application or runtime linker prior to any link-map creation, the following reserved
values for obj are available:
#define PS_OBJ_EXEC ((const char *)0x0) /* application id */
#define PS_OBJ_LDSO ((const char *)0x1) /* runtime linker id */
The controlling process can use the procfs file system for these objects, using the
following pseudo code:
ioctl(.., PIOCNAUXV, ...) - obtain AUX vectors
ldsoaddr = auxv[AT_BASE];
ldsofd = ioctl(..., PIOCOPENM, &ldsoaddr);
Once the file descriptors are found, the ELF files can be examined for their symbol
information by the controlling program.
This chapter describes the executable and linking format (ELF) of the object files
produced by the assembler and link-editor. There are three main types of object files:
A relocatable file holds sections containing code and data. These files are suitable to
be linked with other object files to create executable files, shared object files, or
another relocatable object.
An executable file holds a program that is ready to execute. The file specifies how
exec(2) creates a programs process image.
A shared object file holds code and data suitable to be linked in two contexts. First,
the link-editor can process this file with other relocatable and shared object files to
create other object files. Second, the runtime linker combines this file with a
dynamic executable file and other shared objects to create a process image.
The first section in this chapter, File Format on page 177, focuses on the format of
object files and how that pertains to creating programs. The second section, Dynamic
Linking on page 236, focuses on how the format pertains to loading programs.
Programs manipulate object files with the functions contained in the ELF access
library, libelf. Refer to the elf(3ELF) man page for a description of libelf
contents. Sample source code that uses libelf is provided in the SUNWosdem
package under the /usr/demo/ELF directory.
File Format
Object files participate in both program linking and program execution. For
convenience and efficiency, the object file format provides parallel views of a files
contents, reflecting the differing needs of these activities. The following figure shows
an object files organization.
177
Linking view Execution view
Section 1
Segment 1
...
Section n
Segment 2
...
... ...
An ELF header resides at the beginning of an object file and holds a road map
describing the files organization.
Note Only the ELF header has a fixed position in the file. The flexibility of the ELF
format requires no specified order for header tables, sections or segments. However,
this figure is typical of the layout used in Solaris.
Sections represent the smallest indivisible units that can be processed within an ELF
file. Segments are a collection of sections that represent the smallest individual units
that can be mapped to a memory image by exec(2) or by the runtime linker.
Sections hold the bulk of object file information for the linking view: instructions, data,
symbol table, relocation information, and so on. Descriptions of sections appear in the
first part of this chapter. The second part of this chapter discusses segments and the
program execution view of the file.
A program header table, if present, tells the system how to create a process image.
Files used to generate a process image, executables and shared objects, must have a
program header table; relocatable objects do not need such a table.
Data Representation
The object file format supports various processors with 8-bit bytes, 32bit and 64bit
architectures. Nevertheless, it is intended to be extensible to larger (or smaller)
architectures. Table 71 and Table 72 list the 32bit and 64bit data types.
Object files represent some control data with a machine-independent format. making it
possible to identify object files and interpret their contents in a common way. The
remaining data in an object file use the encoding of the target processor, regardless of
the machine on which the file was created.
ELF Header
Some object file control structures can grow because the ELF header contains their
actual sizes. If the object file format changes, a program can encounter control
structures that are larger or smaller than expected. Programs might therefore ignore
extra information. The treatment of missing information depends on context and will
be specified if and when extensions are defined.
typedef struct {
unsigned char e_ident[EI_NIDENT];
Elf32_Half e_type;
Elf32_Half e_machine;
Elf32_Word e_version;
Elf32_Addr e_entry;
Elf32_Off e_phoff;
Elf32_Off e_shoff;
Elf32_Word e_flags;
Elf32_Half e_ehsize;
Elf32_Half e_phentsize;
Elf32_Half e_phnum;
Elf32_Half e_shentsize;
Elf32_Half e_shnum;
Elf32_Half e_shstrndx;
} Elf32_Ehdr;
typedef struct {
unsigned char e_ident[EI_NIDENT];
Elf64_Half e_type;
Elf64_Half e_machine;
Elf64_Word e_version;
Elf64_Addr e_entry;
Elf64_Off e_phoff;
Elf64_Off e_shoff;
Elf64_Word e_flags;
Elf64_Half e_ehsize;
Although the core file contents are unspecified, type ET_CORE is reserved to mark
the file. Values from ET_LOPROC through ET_HIPROC (inclusive) are reserved for
processor-specific semantics. Other values are reserved and will be assigned to new
object file types as necessary.
e_machine
Specifies the required architecture for an individual file. Relevant architectures are
listed in the following table.
EM_NONE 0 No machine
EM_SPARC 2 SPARC
EM_SPARCV9 43 SPARC V9
Other values are reserved and will be assigned to new machines as necessary (see
sys/elf.h). Processor-specific ELF names use the machine name to distinguish
them. For example, the flags defined in Table 75 use the prefix EF_. A flag named
WIDGET for the EM_XYZ machine would be called EF_XYZ_WIDGET.
e_version
Identifies the object file version, as listed in the following table.
The value 1 signifies the original file format. The value of EV_CURRENT changes as
necessary to reflect the current version number.
e_entry
The virtual address to which the system first transfers control, thus starting the
process. If the file has no associated entry point, this member holds zero.
e_phoff
The program header tables file offset in bytes. If the file has no program header
table, this member holds zero.
e_shoff
The section header tables file offset in bytes. If the file has no section header table,
this member holds zero.
e_flags
Processor-specific flags associated with the file. Flag names take the form
EF_machine_flag. This member is presently zero for x86. The SPARC flags are
listed in the following table.
e_ehsize
The ELF headers size in bytes.
e_phentsize
The size in bytes of one entry in the files program header table. All entries are the
same size.
e_phnum
The number of entries in the program header table. The product of e_phentsize
and e_phnum gives the tables size in bytes. If a file has no program header table,
e_phnum holds the value zero.
e_shentsize
A section headers size in bytes. A section header is one entry in the section header
table. All entries are the same size.
e_shnum
The number of entries in the section header table. The product of e_shentsize
and e_shnum gives the section header tables size in bytes. If a file has no section
header table, e_shnum holds the value zero.
If the section name string table section index is greater than or equal to
SHN_LORESERVE (0xff00), this member has the value SHN_XINDEX (0xffff)
and the actual index of the section name string table section is contained in
thesh_link field of the section header at index 0. Otherwise, the sh_link
member of the initial entry contains 0.
The initial bytes of an ELF header and an object file correspond to the e_ident
member.
These indexes access bytes that hold the values described below.
EI_MAG0 - EI_MAG3
A 4byte magic number, identifying the file as an ELF object file, as listed in the
following table.
ELFMAG1 E e_ident[EI_MAG1]
ELFMAG2 L e_ident[EI_MAG2]
ELFMAG3 F e_ident[EI_MAG3]
EI_CLASS
Byte e_ident[EI_CLASS] identifies the files class, or capacity, as listed in the
following table.
The file format is designed to be portable among machines of various sizes, without
imposing the sizes of the largest machine on the smallest. The class of the file
defines the basic types used by the data structures of the object file container itself.
The data contained in object file sections may follow a different programming
model.
Class ELFCLASS32 supports machines with files and virtual address spaces up to 4
gigabytes. It uses the basic types defined in Table 71.
Class ELFCLASS64 is reserved for 64bit architectures such as SPARC. It uses the
basic types defined in Table 72.
EI_DATA
Byte e_ident[EI_DATA] specifies the data encoding of the processor-specific data
in the object file, as listed in the following table.
Data Encoding
A files data encoding specifies how to interpret the basic objects in a file. Class
ELFCLASS32 files use objects that occupy 1, 2, and 4 bytes. Class ELFCLASS64 files
use objects that occupy 1, 2, 4, and 8 bytes. Under the defined encodings, objects are
represented as shown below. Byte numbers appear in the upper left corners.
0
0x01 01
0 1
0x0102 02 01
0 1 2 3
0x01020304 04 03 02 01
0 1
0x0102 01 02
0 1 2 3
0x01020304 01 02 03 04
Sections
An object files section header table helps you locate all of the sections of the file. The
section header table is an array of Elf32_Shdr or Elf64_Shdr structures, as
described below. A section header table index is a subscript into this array. The ELF
headers e_shoff member gives the byte offset from the beginning of the file to the
section header table; e_shnum tells how many entries the section header table
contains; e_shentsize gives the size in bytes of each entry.
Some section header table indexes are reserved in contexts where index size is
restricted. For example, the st_shndx member of a symbol table entry and the
e_shnum and e_shstrndx members of the ELF header. In such contexts, the reserved
values do not represent actual sections in the object file. Also in such contexts, an
escape value indicates that the actual section index is to be found elsewhere, in a
larger field.
Name Value
SHN_UNDEF 0
SHN_LORESERVE 0xff00
SHN_LOPROC 0xff00
SHN_BEFORE 0xff00
SHN_AFTER 0xff01
SHN_HIPROC 0xff1f
SHN_LOOS 0xff20
SHN_HIOS 0xff3f
SHN_ABS 0xfff1
SHN_COMMON 0xfff2
SHN_XINDEX 0xffff
SHN_HIRESERVE 0xffff
Note Although index 0 is reserved as the undefined value, the section header table
contains an entry for index 0. That is, if the e_shnum member of the ELF header says a
file has 6 entries in the section header table, they have the indexes 0 through 5. The
contents of the initial entry are specified later in this section.
SHN_UNDEF
An undefined, missing, irrelevant, or otherwise meaningless section reference. For
example, a symbol defined relative to section number SHN_UNDEF is an undefined
symbol.
SHN_LORESERVE
The lower boundary of the range of reserved indexes.
SHN_LOPROC - SHN_HIPROC
Values in this inclusive range are reserved for processor-specific semantics.
SHN_LOOS - SHN_HIOS
Values in this inclusive range are reserved for operating system-specific semantics.
SHN_BEFORE, SHN_AFTER
Provide for initial and final section ordering in conjunction with the
SHF_LINK_ORDER and SHF_ORDERED section flags, listed in Table 714.
SHN_ABS
Absolute values for the corresponding reference. For example, symbols defined
relative to section number SHN_ABS have absolute values and are not affected by
relocation.
SHN_COMMON
Symbols defined relative to this section are common symbols, such as FORTRAN
COMMON or unallocated C external variables. These symbols are sometimes referred
to as tentative.
Sections contain all information in an object file except the ELF header, the program
header table, and the section header table. Moreover, the sections in object files satisfy
several conditions:
Every section in an object file has exactly one section header describing it. Section
headers can exist that do not have a section.
Each section occupies one contiguous, possibly empty, sequence of bytes within a
file.
Sections in a file cannot overlap. No byte in a file resides in more than one section.
An object file can have inactive space. The various headers and the sections might
not cover every byte in an object file. The contents of the inactive data are
unspecified.
typedef struct {
Elf64_Word sh_name;
Elf64_Word sh_type;
Elf64_Xword sh_flags;
Elf64_Addr sh_addr;
Elf64_Off sh_offset;
Elf64_Xword sh_size;
Elf64_Word sh_link;
Elf64_Word sh_info;
Elf64_Xword sh_addralign;
Elf64_Xword sh_entsize;
} Elf64_Shdr;
A section headers sh_type member specifies the sections semantics, as shown in the
following table.
Name Value
SHT_NULL 0
SHT_PROGBITS 1
SHT_SYMTAB 2
SHT_STRTAB 3
SHT_RELA 4
SHT_HASH 5
SHT_DYNAMIC 6
SHT_NOTE 7
SHT_NOBITS 8
SHT_REL 9
SHT_SHLIB 10
SHT_DYNSYM 11
SHT_INIT_ARRAY 14
SHT_FINI_ARRAY 15
SHT_PREINIT_ARRAY 16
SHT_GROUP 17
SHT_SYMTAB_SHNDX 18
SHT_LOOS 0x60000000
SHT_LOSUNW 0x6ffffff7
SHT_SUNW_ANNOTATE 0x6ffffff7
SHT_SUNW_DEBUGSTR 0x6ffffff8
SHT_SUNW_DEBUG 0x6ffffff9
SHT_SUNW_move 0x6ffffffa
SHT_SUNW_COMDAT 0x6ffffffb
SHT_SUNW_syminfo 0x6ffffffc
SHT_SUNW_verdef 0x6ffffffd
SHT_SUNW_verneed 0x6ffffffe
SHT_SUNW_versym 0x6fffffff
SHT_HISUNW 0x6fffffff
SHT_HIOS 0x6fffffff
SHT_LOPROC 0x70000000
SHT_SPARC_GOTDATA 0x70000000
SHT_HIPROC 0x7fffffff
SHT_LOUSER 0x80000000
SHT_HIUSER 0xffffffff
SHT_NULL
Identifies the section header as inactive. This section header does not have an
associated section. Other members of the section header have undefined values.
SHT_PROGBITS
Identifies information defined by the program, whose format and meaning are
determined solely by the program.
SHT_SYMTAB, SHT_DYNSYM
Identifies a symbol table. Typically, a SHT_SYMTAB section provides symbols for
link-editing. As a complete symbol table, it can contain many symbols unnecessary
for dynamic linking. Consequently, an object file can also contain a SHT_DYNSYM
section, which holds a minimal set of dynamic linking symbols, to save space. See
Symbol Table Section on page 222 for details.
SHT_STRTAB, SHT_DYNSTR
Identifies a string table. An object file can have multiple string table sections. See
String Table Section on page 221 for details.
SHT_RELA
Identifies relocation entries with explicit addends, such as type Elf32_Rela for the
32bit class of object files. An object file can have multiple relocation sections. See
Relocation Sections on page 210 for details.
SHT_HASH
Identifies a symbol hash table. A dynamically linked object file must contain a
symbol hash table. Currently, an object file can have only one hash table, but this
restriction might be relaxed in the future. See Hash Table Section on page 205 for
details.
SHT_DYNAMIC
Identifies information for dynamic linking. Currently, an object file can have only
one dynamic section. See Dynamic Section on page 248 for details.
Other section-type values are reserved. As mentioned before, the section header for
index 0 (SHN_UNDEF) exists, even though the index marks undefined section
references. The following table shows the values.
sh_name 0 No name
sh_flags 0 No flags
sh_addr 0 No address
sh_size 0 No size
sh_addralign 0 No alignment
sh_entsize 0 No entries
A section headers sh_flags member holds 1-bit flags that describe the sections
attributes:
Name Value
SHF_WRITE 0x1
SHF_ALLOC 0x2
SHF_EXECINSTR 0x4
SHF_MERGE 0x10
SHF_STRINGS 0x20
SHF_INFO_LINK 0x40
SHF_LINK_ORDER 0x80
SHF_OS_NONCONFORMING 0x100
SHF_GROUP 0x200
SHF_TLS 0x400
SHF_MASKOS 0x0ff00000
SHF_ORDERED 0x40000000
SHF_EXCLUDE 0x80000000
SHF_MASKPROC 0xf0000000
If a flag bit is set in sh_flags, the attribute is on for the section. Otherwise, the
attribute is off or does not apply. Undefined attributes are reserved and set to zero.
SHF_WRITE
Identifies a section that should be writable during process execution.
SHF_ALLOC
Identifies a section that occupies memory during process execution. Some control
sections do not reside in the memory image of an object file. This attribute is off for
those sections.
SHF_EXECINSTR
Identifies a section that contains executable machine instructions.
SHF_MERGE
Identifies a section containing data that may be merged to eliminate duplication.
Unless the SHF_STRINGS flag is also set, the data elements in the section are of a
uniform size. The size of each element is specified in the section headers
sh_entsize field. If the SHF_STRINGS flag is also set, the data elements consist of
null-terminated character strings. The size of each character is specified in the
section headers sh_entsize field.
SHF_STRINGS
Identifies a section that consists of null-terminated character strings. The size of
each character is specified in the section headers sh_entsize field.
SHF_INFO_LINK
This section headers sh_info field holds a section header table index.
SHF_LINK_ORDER
This section adds special ordering requirements to the link-editor. The requirements
apply if the sh_link field of this sections header references another section, the
linked-to section. If this section is combined with other sections in the output file,
the section appears in the same relative order with respect to those sections.
Similarly the linked-to section appears with respect to sections the linked-to section
is combined with.
A typical use of this flag is to build a table that references text or data sections in
address order.
In the absence of the sh_link ordering information, sections from a single input
file combined within one section of the output file will be contiguous and have the
same relative ordering as they did in the input file. The contributions from multiple
input files appear in link-line order.
SHF_OS_NONCONFORMING
This section requires special OS-specific processing beyond the standard linking
rules to avoid incorrect behavior. If this section has either an sh_type value or
contains sh_flags bits in the OS-specific ranges for those fields, and the
link-editor does not recognize these values, then the link-editor will reject the object
file containing this section with an error.
SHF_GROUP
This section is a member, perhaps the only one, of a section group. The section must
be referenced by a section of type SHT_GROUP. The SHF_GROUP flag can be set only
for sections contained in relocatable objects. See Group Section on page 204 for
details.
SHF_TLS
This section holds thread-local storage, meaning that each separate execution flow
has its own distinct instance of this data. See Chapter 8 for details.
SHF_MASKOS
All bits included in this mask are reserved for operating system-specific semantics.
SHF_ORDERED
This section requires ordering in relation to other sections of the same type. Ordered
sections are combined within the section pointed to by the sh_link entry. The
sh_link entry of an ordered section can point to itself.
If the sh_info entry of the ordered section is a valid section within the same input
file, the ordered section will be sorted based on the relative ordering within the
output file of the section pointed to by the sh_info entry.
The special sh_info values SHN_BEFORE and SHN_AFTER (see Table 711) imply
that the sorted section is to precede or follow, respectively, all other sections in the
set being ordered. Input file link-line order is preserved if multiple sections in an
ordered set have one of these special values.
In the absence of the sh_info ordering information, sections from a single input
file combined within one section of the output file will be contiguous and have the
same relative ordering as they did in the input file. The contributions from multiple
input files appear in link-line order.
Two members in the section header, sh_link and sh_info, hold special information,
depending on section type.
SHT_REL The section header index of the The section header index of the
associated symbol table. section to which the relocation
SHT_RELA
applies. See also Table 716 and
Relocation Sections
on page 210.
SHT_SYMTAB The section header index of the One greater than the symbol
associated string table. table index of the last local
SHT_DYNSYM
symbol (binding STB_LOCAL).
SHT_GROUP The section header index of the The symbol table index of an
associated symbol table. entry in the associated symbol
table. The name of the specified
symbol table entry provides a
signature for the section group.
SHT_SUNW_COMDAT 0 0
SHT_SUNW_syminfo The section header index of the The section header index of the
associated symbol table. associated .dynamic section.
Special Sections
Various sections hold program and control information. Sections in the following table
are used by the system and have the indicated types and attributes.
.bss
Uninitialized data that contribute to the programs memory image. By definition,
the system initializes the data with zeros when the program begins to run. The
section occupies no file space, as indicated by the section type SHT_NOBITS.
Section names with a dot (.) prefix are reserved for the system, although applications
can use these sections if their existing meanings are satisfactory. Applications can use
names without the prefix to avoid conflicts with system sections. The object file format
enables you to define sections not in the list above. An object file can have more than
one section with the same name.
COMDAT Section
COMDAT sections are uniquely identified by their section name (sh_name). If the
link-editor encounters multiple sections of type SHT_SUNW_COMDAT, with the same
section name, the first section is retained and the rest discarded. Any relocations
applied to a discarded SHT_SUNW_COMDAT section are ignored. Any symbols defined
in a discarded section are removed.
Group Section
Some sections occur in interrelated groups. For example, an out-of-line definition of an
inline function might require additional information besides the section containing
executable instructions. This additional information can be a read-only data section
containing literals referenced, one or more debugging information sections, or other
informational sections. Furthermore, there may be internal references among these
sections. These references make no sense if one of the sections were removed, or one of
the sections were replaced, by a duplicate from another object. Therefore, these groups
are included, or these groups are omitted, from the linked object as a unit.
The section data of a SHT_GROUP section is an array of Elf32_Word entries. The first
entry is a flag word. The remaining entries are a sequence of section header indices.
Name Value
GRP_COMDAT 0x1
GRP_COMDAT
GRP_COMDAT is a COMDAT group. It may duplicate another COMDAT group in
another object file, where duplication is defined as having the same group
signature. In such cases, only one of the duplicate groups is retained by the
link-editor. The members of the remaining groups are discarded.
The section header indices in the SHT_GROUP section, identify the sections that make
up the group. These sections must have the SHF_GROUP flag set in their sh_flags
section header member. If the link-editor decides to remove the section group, the
link-editor removes all members of the group.
nbucket
nchain
bucket [0]
...
bucket [nbucket-1]
chain [0]
...
chain [nchain-1]
The bucket array contains nbucket entries, and the chain array contains nchain
entries. Indexes start at 0. Both bucket and chain hold symbol table indexes. Chain
table entries parallel the symbol table. The number of symbol table entries should
equal nchain, so symbol table indexes also select chain table entries.
The chain links can be followed until the selected symbol table entry holds the
desired name, or the chain entry contains the value STN_UNDEF.
while (*name)
{
h = (h << 4) + *name++;
if (g = h & 0xf0000000)
h ^= g >> 24;
h &= ~g;
}
return h;
}
Move Section
Typically, within ELF files, initialized data variables are maintained within the object
file. If a data variable is very large, and contains only a small number of initialized
(nonzero) elements, the entire variable is still maintained in the object file.
Objects that contain large partially initialized data variables, such as FORTRAN
COMMON blocks, can result in a significant disk space overhead. The SHT_SUNW_move
section provides a mechanism of compressing these data variables. This compression
reduces the disk size of the associated object.
typedef struct {
Elf64_Lword m_value;
Elf64_Xword m_info;
Elf64_Xword m_poffset;
Elf64_Half m_repeat;
Elf64_Half m_stride;
} Elf64_Move;
The following data definition would traditionally consume 0x8000 bytes within an
object file:
typedef struct {
int one;
char two;
} Data
Data move[0x1000] = {
{0, 0}, {1, 1}, {0, 0},
{0xf, F}, {0xf, F}, {0, 0},
{0xe, E}, {0, 0}, {0xe, E}
};
Move sections supplied from relocatable objects are concatenated and output in the
object being created by the link-editor. However, the following conditions cause the
link-editor to process the move entries and expand their contents into a traditional
data item:
The output file is a static executable.
The size of the move entries is greater than the size of the symbol into which the
move data would be expanded.
The -z nopartial option is in effect.
Note Section
Sometimes a vendor or system engineer needs to mark an object file with special
information that other programs will check for conformance, compatibility, and so
forth. Sections of type SHT_NOTE and program header elements of type PT_NOTE can
be used for this purpose.
The note information in sections and program header elements holds any number of
entries, as shown in the following figure. For 64 and 32bit objects, each entry is an
array of 4-byte words in the format of the target processor. Labels are shown in Figure
76 to help explain note information organization, but they are not part of the
specification.
descsz
type
name
...
desc
...
The note segment shown in the following figure holds two entries.
descsz 0 No descriptor
type 1
name X Y Z
C o \0 pad
namesz 7
descsz 8
type 3
name X Y Z
C o \0 pad
desc word0
word1
Note The system reserves note information with no name (namesz == 0) and with a
zero-length name (name[0] == \0) but currently defines no types. All other
names must have at least one non-null character.
Relocation Sections
Relocation is the process of connecting symbolic references with symbolic definitions.
For example, when a program calls a function, the associated call instruction must
transfer control to the proper destination address at execution. Relocatable files must
have information that describes how to modify their section contents, thus allowing
executable and shared object files to hold the right information for a processs program
image. Relocation entries are these data.
typedef struct {
Elf32_Addr r_offset;
typedef struct {
Elf64_Addr r_offset;
Elf64_Xword r_info;
} Elf64_Rel;
typedef struct {
Elf64_Addr r_offset;
Elf64_Xword r_info;
Elf64_Sxword r_addend;
} Elf64_Rela;
For a relocatable file, the value indicates a section offset. The relocation section itself
describes how to modify another section in the file. Relocation offsets designate a
storage unit within the second section.
For an executable or shared object, the value indicates the virtual address of the
storage unit affected by the relocation. This information makes the relocation
entries more useful for the runtime linker.
Although the interpretation of the member changes for different object files to allow
efficient access by the relevant programs, the meanings of the relocation types stay
the same.
r_info
This member gives both the symbol table index, with respect to which the
relocation must be made, and the type of relocation to apply. For example, a call
instructions relocation entry holds the symbol table index of the function being
called. If the index is STN_UNDEF, the undefined symbol index, the relocation uses
0 as the symbol value.
r_addend
This member specifies a constant addend used to compute the value to be stored
into the relocatable field.
Rela entries contain an explicit addend. Entries of type Rel store an implicit addend
in the location to be modified. 32bit and 64bit SPARC use only Elf32_Rela and
Elf64_Rela relocation entries respectively. Thus, the r_addend member serves as
the relocation addend. x86 uses only Elf32_Rel relocation entries. The field to be
relocated holds the addend. In all cases, the addend and the computed result use the
same byte order.
A relocation section can reference two other sections: a symbol table, identified by the
sh_info section header entry, and a section to modify, identified by the sh_link
section header entry. Sections on page 187 specifies these relationships. An sh_link
entry is required when a relocation section exists in a relocatable object, but is optional
for executables and shared objects. The relocation offset is sufficient to perform the
relocation.
half16
15 0
word32
31 0
disp30
31 29 0
disp22
31 21 0
imm22
31 21 0
disp19
31 19 0
d2 disp14
31 21 19 13 0
simm13
31 12 0
simm11
31 10 0
simm10
31 9 0
simm7
31 6 0
xword64
63 0
word32
31 0
word32 specifies a 32bit field occupying 4 bytes with an arbitrary byte alignment.
These values use the same byte order as other word values in the x86 architecture:
3 2 1 0
01 02 03 04 0x01020304
31 0
In all cases, the r_offset value designates the offset or virtual address of the first
byte of the affected storage unit. The relocation type specifies which bits to change and
how to calculate their values.
Calculations for the following relocation types assume the actions are transforming a
relocatable file into either an executable or a shared object file. Conceptually, the
link-editor merges one or more relocatable files to form the output. The link-editor
first decides how to combine and locate the input files. Then it updates the symbol
values. Finally the link-editor performs the relocation. Relocations applied to
executable or shared object files are similar and accomplish the same result.
Descriptions in the tables in this section use the following notation:
A The addend used to compute the value of the relocatable field.
B The base address at which a shared object is loaded into memory during
execution. Generally, a shared object file is built with a 0 base virtual address,
but the execution address is different. See Program Header on page 236.
G The offset into the global offset table at which the address of the relocation
entrys symbol resides during execution. See Global Offset Table
(Processor-Specific) on page 260.
GOT The address of the global offset table. See Global Offset Table
(Processor-Specific) on page 260.
L The section offset or address of the procedure linkage table entry for a symbol.
See Procedure Linkage Table (Processor-Specific) on page 261.
P The section offset or address of the storage unit being relocated, computed
using r_offset.
S The value of the symbol whose index resides in the relocation entry.
R_SPARC_8 1 V-byte8 S + A
R_SPARC_16 2 V-half16 S + A
R_SPARC_32 3 V-word32 S + A
R_SPARC_DISP8 4 V-byte8 S + A - P
R_SPARC_DISP16 5 V-half16 S + A - P
R_SPARC_DISP32 6 V-disp32 S + A - P
R_SPARC_22 10 V-imm22 S + A
R_SPARC_13 11 V-simm13 S + A
R_SPARC_GOT13 14 V-simm13 G
R_SPARC_GLOB_DAT 20 V-word32 S + A
R_SPARC_RELATIVE 22 V-word32 B + A
R_SPARC_UA32 23 V-word32 S + A
R_SPARC_PLT32 24 V-word32 L + A
R_SPARC_PCPLT32 27 V-word32 L + A - P
R_SPARC_10 30 V-simm10 S + A
R_SPARC_11 31 V-simm11 S + A
R_SPARC_7 43 V-imm7 S + A
R_SPARC_5 44 V-imm5 S + A
R_SPARC_6 45 V-imm6 S + A
R_SPARC_REGISTER 53 V-word32 S + A
R_SPARC_UA16 55 V-half16 S + A
Note Additional relocations are available for thread-local storage references. These
relocations are covered in Chapter 8.
The relocations listed in the following table extend, or alter, those define for 32bit
SPARC. See SPARC: Relocation Types on page 215.
R_SPARC_GLOB_DAT 20 V-xword64 S + A
R_SPARC_RELATIVE 22 V-xword64 B + A
R_SPARC_64 32 V-xword64 S + A
R_SPARC_DISP64 46 V-xword64 S + A - P
R_SPARC_PLT64 47 V-xword64 L + A
R_SPARC_REGISTER 53 V-xword64 S + A
R_SPARC_UA64 54 V-xword64 S + A
R_386_32 1 word32 S + A
R_386_PC32 2 word32 S + A - P
R_386_GOT32 3 word32 G + A
R_386_PLT32 4 word32 L + A - P
R_386_GLOB_DAT 6 word32 S
R_386_JMP_SLOT 7 word32 S
R_386_RELATIVE 8 word32 B + A
R_386_32PLT 11 word32 L + A
Note Additional relocations are available for thread-local storage references. These
relocations are covered in Chapter 8.
The first byte, which is index zero, holds a null character. Likewise, a string tables last
byte holds a null character, ensuring null termination for all strings. A string whose
index is zero specifies either no name or a null name, depending on the context.
An empty string table section is permitted. The section headers sh_size member
contains zero. Nonzero indexes are invalid for an empty string table.
A section headers sh_name member holds an index into the section header string
table section, as designated by the e_shstrndx member of the ELF header. The
following figure shows a string table with 25 bytes and the strings associated with
various indexes.
The table below shows the strings of the string table shown in the preceding figure.
Index String
0 None
1 name
7 Variable
11 able
16 able
24 null string
As the example shows, a string table index can refer to any byte in the section. A string
can appear more than once. References to substrings can exist. A single string can be
referenced multiple times. Unreferenced strings also are allowed.
typedef struct {
Elf64_Word st_name;
st_other
A symbols visibility. A list of the values and meanings appears in Table 724. The
following code shows how to manipulate the values for both 32bit and 64bit
objects. Other bits contain 0 and have no defined meaning.
#define ELF32_ST_VISIBILITY(o) ((o)&0x3)
#define ELF64_ST_VISIBILITY(o) ((o)&0x3)
st_shndx
Every symbol table entry is defined in relation to some section. This member holds
the relevant section header table index. Some section indexes indicate special
meanings. See Table 711.
If this member contains SHN_XINDEX, then the actual section header index is too
large to fit in this field. The actual value is contained in the associated section of
type SHT_SYMTAB_SHNDX.
Name Value
STB_LOCAL 0
STB_GLOBAL 1
STB_WEAK 2
STB_LOOS 10
STB_HIOS 12
STB_LOPROC 13
STB_HIPROC 15
STB_LOCAL
Local symbol. These symbols are not visible outside the object file containing their
definition. Local symbols of the same name can exist in multiple files without
interfering with each other.
STB_GLOBAL
Global symbols. These symbols are visible to all object files being combined. One
files definition of a global symbol satisfies another files undefined reference to the
same global symbol.
STB_WEAK
Weak symbols. These symbols resemble global symbols, but their definitions have
lower precedence.
STB_LOOS - STB_HIOS
Values in this inclusive range are reserved for operating system-specific semantics.
STB_LOPROC - STB_HIPROC
Values in this inclusive range are reserved for processor-specific semantics.
Note Weak symbols are intended primarily for use in system software. Their use in
application programs is discouraged.
In each symbol table, all symbols with STB_LOCAL binding precede the weak and
global symbols. As Sections on page 187 describes, a symbol table sections
sh_info section header member holds the symbol table index for the first non-local
symbol.
A symbols type, determined from its st_info field, provides a general classification
for the associated entity.
Name Value
STT_NOTYPE 0
STT_OBJECT 1
STT_FUNC 2
STT_SECTION 3
STT_FILE 4
STT_COMMON 5
STT_TLS 6
STT_LOOS 10
STT_HIOS 12
STT_LOPROC 13
STT_SPARC_REGISTER 13
STT_HIPROC 15
STT_NOTYPE
The symbol type is not specified.
Name Value
STV_DEFAULT 0
STV_INTERNAL 1
STV_HIDDEN 2
STV_PROTECTED 3
None of the visibility attributes affects the resolution of symbols within an executable
or shared object during link-editing. Such resolution is controlled by the binding type.
Once the link-editor has chosen its resolution, these attributes impose two
requirements. Both requirements are based on the fact that references in the code being
linked may have been optimized to take advantage of the attributes.
First, all of the non-default visibility attributes, when applied to a symbol
reference, imply that a definition to satisfy that reference must be provided within
the current executable or shared object. If this type of symbol reference has no
definition within the component being linked, then the reference must have
STB_WEAK binding and is resolved to zero.
Second, if any reference to or definition of a name is a symbol with a non-default
visibility attribute, the visibility attribute must be propagated to the resolving
symbol in the linked object. If different visibility attributes are specified for distinct
references to or definitions of a symbol, the most constraining visibility attribute
must be propagated to the resolving symbol in the linked object. The attributes,
ordered from least to most constraining, are STV_PROTECTED, STV_HIDDEN and
STV_INTERNAL.
As mentioned above, the symbol table entry for index 0 (STN_UNDEF) is reserved. This
entry holds the values listed in the following table.
TABLE 725 ELF Symbol Table Entry: Index 0
st_name 0 No name
st_size 0 No size
st_other 0
Symbol Values
Symbol table entries for different object file types have slightly different interpretations
for the st_value member.
In relocatable files, st_value holds alignment constraints for a symbol whose
section index is SHN_COMMON.
In relocatable files, st_value holds a section offset for a defined symbol.
st_value is an offset from the beginning of the section that st_shndx identifies.
In executable and shared object files, st_value holds a virtual address. To make
these files symbols more useful for the runtime linker, the section offset (file
interpretation) gives way to a virtual address (memory interpretation) for which
Although the symbol table values have similar meanings for different object files, the
data allow efficient access by the appropriate programs.
Register Symbols
The SPARC architecture supports register symbols, which are symbols that initialize a
global register. A symbol table entry for a register symbol contains the entries listed in
the following table.
TABLE 726 SPARC: ELF Symbol Table Entry: Register Symbol
Field Meaning
st_name Index into string table of the name of the symbol, or 0 for a
scratch register.
st_value Register number. See the ABI manual for integer register
assignments.
The register values defined for SPARC are listed in the following table.
TABLE 727 SPARC: ELF Register Numbers
Absence of an entry for a particular global register means that the particular global
register is not used at all by the object.
Index 0 is used to store the current version of the Syminfo table, which is
SYMINFO_CURRENT. Since symbol table entry 0 is always reserved for the UNDEF
symbol table entry, this does not pose any conflicts.
typedef struct {
Elf64_Half si_boundto;
Elf64_Half si_flags;
} Elf64_Syminfo;
si_flags
This bit-field can have flags set, as shown in the following table.
Versioning Sections
Objects created by the link-editor can contain two types of versioning information:
Version definitions provide associations of global symbols and are implemented
using sections of type SHT_SUNW_verdef and SHT_SUNW_versym.
Version dependencies indicate the version definition requirements from other object
dependencies and are implemented using sections of type SHT_SUNW_verneed.
The structures that form these sections are defined in sys/link.h. Sections that
contain versioning information are named .SUNW_version.
typedef struct {
Elf32_Word vda_name;
Elf32_Word vda_next;
} Elf32_Verdaux;
typedef struct {
Elf64_Word vda_name;
Elf64_Word vda_next;
} Elf64_Verdaux;
The value 1 signifies the original section format. Extensions will create new
versions with higher numbers. The value of VER_DEF_CURRENT changes as
necessary to reflect the current version number.
vd_flags
This member holds version definition-specific information, as listed in the
following table.
TABLE 731 ELF Version Definition Section Flags
The base version definition is always present when version definitions, or symbol
auto-reduction, have been applied to the file. The base version provides a default
version for the files reserved symbols. A weak version definition has no symbols
associated with it. See Creating a Weak Version Definition on page 134.
vd_ndx
The version index. Each version definition has a unique index that is used to
associate SHT_SUNW_versym entries to the appropriate version definition.
The number of elements of the array must equal the number of symbol table entries
contained in the associated symbol table. This number is determined by the sections
sh_link value. Each element of the array contains a single index that can have the
values shown in the following table.
Any index values greater than VER_NDX_GLOBAL must correspond to the vd_ndx
value of an entry in the SHT_SUNW_verdef section. If no index values greater than
VER_NDX_GLOBAL exist, then no SHT_SUNW_verdef section need be present.
typedef struct {
Elf32_Word vna_hash;
Elf32_Half vna_flags;
Elf32_Half vna_other;
Elf32_Word vna_name;
Elf32_Word vna_next;
} Elf32_Vernaux;
typedef struct {
Elf64_Half vn_version;
Elf64_Half vn_cnt;
Elf64_Word vn_file;
Elf64_Word vn_aux;
Elf64_Word vn_next;
} Elf64_Verneed;
typedef struct {
Elf64_Word vna_hash;
Elf64_Half vna_flags;
Elf64_Half vna_other;
Elf64_Word vna_name;
Elf64_Word vna_next;
} Elf64_Vernaux;
The value 1 signifies the original section format. Extensions will create new
versions with higher numbers. The value of VER_NEED_CURRENT changes as
necessary to reflect the current version number.
vn_cnt
The number of elements in the Elf32_Vernaux array.
vn_file
The string table offset to a null-terminated string, that provides the file name
having a version dependency. This name matches one of the .dynamic
dependencies found in the file. See Dynamic Section on page 248.
vn_aux
The byte offset, from the start of this Elf32_Verneed entry, to the
Elf32_Vernaux array of version definitions required from the associated file
dependency. There must exist at least one version dependency. Additional version
dependencies can be present, the number being indicated by the vn_cnt value.
vn_next
The byte offset, from the start of this Elf32_Verneed entry, to the next
Elf32_Verneed entry.
vna_hash
The hash value of the version dependency name. This value is generated using the
same hashing function described in Hash Table Section on page 205.
vna_flags
Version dependency specific information, as listed in the following table.
Dynamic Linking
This section describes the object file information and system actions that create
running programs. Most information here applies to all systems. Information specific
to one processor resides in sections marked accordingly.
Executable and shared object files statically represent application programs. To execute
such programs, the system uses the files to create dynamic program representations,
or process images. A process image has segments that contain its text, data, stack, and
so on. The major subsections of this section are:
Program Header on page 236 describes object file structures that are directly
involved in program execution. The primary data structure, a program header
table, locates segment images in the file and contains other information needed to
create the memory image of the program.
Program Loading (Processor-Specific) on page 241 describes the information
used to load a program into memory.
Runtime Linker on page 247 describes the information used to specify and
resolve symbolic references among the object files of the process image.
Program Header
An executable or shared object files program header table is an array of structures,
each describing a segment or other information that the system needs to prepare the
program for execution. An object file segment contains one or more sections, as
described in Segment Contents on page 241.
Program headers are meaningful only for executable and shared object files. A file
specifies its own program header size with the ELF headers e_phentsize and
e_phnum members..
typedef struct {
Elf64_Word p_type;
Elf64_Word p_flags;
Elf64_Off p_offset;
Elf64_Addr p_vaddr;
Elf64_Addr p_paddr;
Elf64_Xword p_filesz;
Elf64_Xword p_memsz;
Elf64_Xword p_align;
} Elf64_Phdr;
Name Value
PT_NULL 0
PT_LOAD 1
PT_DYNAMIC 2
PT_INTERP 3
PT_NOTE 4
PT_SHLIB 5
PT_PHDR 6
PT_TLS 7
PT_LOSUNW 0x6ffffffa
PT_SUNWBSS 0x6ffffffa
PT_SUNWSTACK 0x6ffffffb
PT_HISUNW 0x6fffffff
PT_LOPROC 0x70000000
PT_HIPROC 0x7fffffff
PT_NULL
Unused; other members values are undefined. This type enables the program
header table to contain ignored entries.
PT_LOAD
Specifies a loadable segment, described by p_filesz and p_memsz. The bytes
from the file are mapped to the beginning of the memory segment. If the segments
memory size (p_memsz) is larger than the file size (p_filesz), the extra bytes are
defined to hold the value 0 and to follow the segments initialized area. The file size
can not be larger than the memory size. Loadable segment entries in the program
header table appear in ascending order, sorted on the p_vaddr member.
PT_DYNAMIC
Specifies dynamic linking information. See Dynamic Section on page 248.
PT_INTERP
Specifies the location and size of a null-terminated path name to invoke as an
interpreter. This segment type is mandatory for dynamic executable files and can
Note Unless specifically required elsewhere, all program header segment types are
optional. A files program header table can contain only those elements relevant to its
contents.
Base Address
Executable and shared object files have a base address, which is the lowest virtual
address associated with the memory image of the programs object file. One use of the
base address is to relocate the memory image of the program during dynamic linking.
To compute the base address, you determine the memory address associated with the
lowest p_vaddr value for a PT_LOAD segment. You then obtain the base address by
truncating the memory address to the nearest multiple of the maximum page size.
Depending on the kind of file being loaded into memory, the memory address might
not match the p_vaddr values.
Segment Permissions
A program to be loaded by the system must have at least one loadable segment,
although this is not required by the file format. When the system creates loadable
segment memory images, it gives access permissions, as specified in the p_flags
member. All bits included in the PF_MASKPROC mask are reserved for
processor-specific semantics.
If a permission bit is 0, that bits type of access is denied. Actual memory permissions
depend on the memory management unit, which can vary from one system to another.
Although all flag combinations are valid, the system can grant more access than
requested. In no case, however, will a segment have write permission unless it is
specified explicitly. The following table lists both the exact flag interpretation and the
allowable flag interpretation.
For example, typical text segments have read and execute, but not write permissions.
Data segments normally have read, write, and execute permissions.
Segment Contents
An object file segment consists of one or more sections, though this fact is transparent
to the program header. Whether the file segment holds one or many sections also is
immaterial to program loading. Nonetheless, various data must be present for
program execution, dynamic linking, and so on. The diagrams below illustrate
segment contents in general terms. The order and membership of sections within a
segment can vary.
Text segments contain read-only instructions and data. Data segments contain writable
data and instructions. See Table 716 for a list of all special sections.
A PT_DYNAMIC program header element points at the .dynamic section. The .got
and .plt sections also hold information related to position-independent code and
dynamic linking.
The .plt can reside in a text or a data segment, depending on the processor. See
Global Offset Table (Processor-Specific) on page 260 and Procedure Linkage Table
(Processor-Specific) on page 261 for details.
The .bss section has the type SHT_NOBITS. Although it occupies no space in the file,
it contributes to the segments memory image. Normally, these uninitialized data
reside at the end of the segment, thereby making p_memsz larger than p_filesz in
the associated program header element.
Virtual addresses and file offsets for 32bit segments are congruent modulo 64K
(0x10000). Virtual addresses and file offsets for 64bit segments are congruent
modulo 1 megabyte (0x100000). By aligning segments to the maximum page size, the
files are suitable for paging regardless of physical page size.
The following figure presents the SPARC version of the executable file.
...
0x4f5 bytes
0x44f5 Other information 0x244f5
...
The following table defines the loadable segment elements for the previous figure.
The following figure presents the x86 version of the executable file.
...
0x3a0 bytes
0x43a0 Other information 0x80643a0
...
The following table defines the loadable segment elements for the previous figure.
The examples file offsets and virtual addresses are congruent modulo the maximum
page size for both text and data. Up to four file pages hold impure text or data
depending on page size and file system block size.
The first text page contains the ELF header, the program header table, and other
information.
The last text page holds a copy of the beginning of data.
The first data page has a copy of the end of text.
The last data page can contain file information not relevant to the running process.
Logically, the system enforces the memory permissions as if each segment were
complete and separate The segments addresses are adjusted to ensure that each
logical page in the address space has a single set of permissions. In the examples
above, the region of the file holding the end of text and the beginning of data is
mapped twice: at one virtual address for text and at a different virtual address for
data.
Note The examples above reflect typical Solaris system binaries that have their text
segments rounded.
The end of the data segment requires special handling for uninitialized data, which
the system defines to begin with zero values. If a files last data page includes
information not in the logical memory page, the extraneous data must be set to zero,
not the unknown contents of the executable file.
Impurities in the other three pages are not logically part of the process image. Whether
the system expunges these impurities is unspecified. The memory image for this
program is shown in the following figures, assuming 4 Kbyte (0x1000) pages. For
simplicity, these figures illustrate only one page size.
...
Text
0x13a82
0x3a82 bytes
Data padding
0x57e
... Data
0x244f5
0x4f5 bytes
Uninitialized data
0x250a4 0xbaf
Page padding
0xaf5c
...
Text
0x80532fd
0x32fd bytes
Data padding
0xd03
... Data
0x80643a0
3a0 bytes
Uninitialized data
0x8064dc4 0xa24
Page padding
0xb23c
One aspect of segment loading differs between executable files and shared objects.
Executable file segments typically contain absolute code. For the process to execute
correctly, the segments must reside at the virtual addresses used to create the
executable file. The system uses the p_vaddr values unchanged as virtual addresses.
Though the system chooses virtual addresses for individual processes, it maintains the
relative positions of the segments. Because position-independent code uses relative
addressing between segments, the difference between virtual addresses in memory
must match the difference between virtual addresses in the file.
Program Interpreter
A dynamic executable or shared object that initiates dynamic linking can have one
PT_INTERP program header element. During exec(2), the system retrieves a path
name from the PT_INTERP segment and creates the initial process image from the
interpreter files segments. The interpreter is responsible for receiving control from the
system and providing an environment for the application program.
In the Solaris operating environment the interpreter is known as the runtime linker,
ld.so.1(1).
Runtime Linker
When creating a dynamic object that initiates dynamic linking, the link-editor adds a
program header element of type PT_INTERP to an executable file. This element
instructing the system to invoke the runtime linker as the program interpreter.
exec(2) and the runtime linker cooperate to create the process image for the program.
Shared objects can occupy virtual memory addresses that are different from the
addresses recorded in the files program header table. The runtime linker relocates the
memory image, updating absolute addresses before the application gains control.
Dynamic Section
If an object file participates in dynamic linking, its program header table will have an
element of type PT_DYNAMIC. This segment contains the .dynamic section. A special
symbol, _DYNAMIC, labels the section, which contains an array of the following
structures, defined in sys/link.h:
typedef struct {
Elf32_Sword d_tag;
union {
Elf32_Word d_val;
Elf32_Addr d_ptr;
Elf32_Off d_off;
} d_un;
} Elf32_Dyn;
typedef struct {
Elf64_Xword d_tag;
union {
Elf64_Xword d_val;
Elf64_Addr d_ptr;
} d_un;
} Elf64_Dyn;
For each object with this type, d_tag controls the interpretation of d_un.
d_val
These objects represent integer values with various interpretations.
d_ptr
These objects represent program virtual addresses. A files virtual addresses might
not match the memory virtual addresses during execution. When interpreting
addresses contained in the dynamic structure, the runtime linker computes actual
To make interpreting the contents of dynamic section entries simpler for tools, the
value of each tag, except for those in two special compatibility ranges, will determine
the interpretation of the d_un union. A tag whose value is an even number indicates a
dynamic section entry that uses d_ptr. A tag whose value is an odd number indicates
a dynamic section entry that uses d_val or that uses neither d_ptr nor d_val. Tags
whose values are less than the special value DT_ENCODING and tags whose values fall
between DT_HIOS and DT_LOPROC do not follow these rules.
The following table summarizes the tag requirements for executable and shared object
files. If a tag is marked mandatory, then the dynamic linking array must have an entry
of that type. Likewise, optional means an entry for the tag can appear but is not
required.
DT_NULL
Marks the end of the _DYNAMIC array.
DT_NEEDED
The DT_STRTAB string table offset of a null-terminated string, giving the name of a
needed dependency. The dynamic array can contain multiple entries of this type.
The relative order of these entries is significant, though their relation to entries of
other types is not. See Shared Object Dependencies on page 64.
An object file can have multiple relocation sections. When creating the relocation
table for an executable or shared object file, the link-editor catenates those sections
to form a single table. Although the sections may remain independent in the object
file, the runtime linker sees a single table. When the runtime linker creates the
process image for an executable file or adds a shared object to the process image, it
reads the relocation table and performs the associated actions.
This element requires the DT_RELASZ and DT_RELAENT elements also be present.
When relocation is mandatory for a file, either DT_RELA or DT_REL can occur.
DT_RELASZ
The total size, in bytes, of the DT_RELA relocation table.
DT_RELAENT
The size, in bytes, of the DT_RELA relocation entry.
DT_STRSZ
The total size, in bytes, of the DT_STRTAB string table.
DT_SYMENT
The size, in bytes, of the DT_SYMTAB symbol entry.
DT_INIT
The address of an initialization function. See Initialization and Termination
Sections on page 36.
DT_FINI
The address of a termination function. See Initialization and Termination Sections
on page 36.
Except for the DT_NULL element at the end of the dynamic array and the relative order
of DT_NEEDED and DT_POSFLAG_1 elements, entries can appear in any order. Tag
values not appearing in the table are reserved.
DF_1_NOW
Indicates that all relocations for this object must be processed before returning
control to the program. The presence of this flag takes precedence over a directive
to use lazy binding when specified through the environment or via dlopen(3DL).
See When Relocations Are Performed on page 71.
DF_1_GROUP
Indicates that the object is a member of a group. This flag is recorded in the object
using the link-editors -B group option. See Object Hierarchies on page 88.
DF_1_NODELETE
Indicates that the object cannot be deleted from a process. If the object is loaded in a
process, either directly or as a dependency, with dlopen(3DL), it cannot be
unloaded with dlclose(3DL). This flag is recorded in the object using the
link-editors -z nodelete option.
DF_1_LOADFLTR
Meaningful only for filters. Indicates that all associated filtees be processed
immediately. This flag is recorded in the object using the link-editors
-z loadfltr option. See Filtee Processing on page 111.
DF_1_INITFIRST
Indicates that this objects initialization section be run before any other objects
loaded with it. This flag is intended for specialized system libraries only, and is
recorded in the object using the link-editors -z initfirst option.
DF_1_NOOPEN
Indicates that the object cannot be added to a running process with dlopen(3DL).
This flag is recorded in the object using the link-editors -z nodlopen option.
DF_1_ORIGIN
Indicates that the object requires $ORIGIN processing. See Locating Associated
Dependencies on page 323.
DF_1_DIRECT
Indicates that the object should use direct binding information. See Direct
Binding on page 69.
DTF_1_PARINIT
Indicates that the object requires partial initialization. See Move Section
on page 206.
DTF_1_CONFEXP
Identifies this object as a configuration alternative object generated by crle(1). This
flag triggers the runtime linker to search for a configuration file
$ORIGIN/ld.config.app-name. This flag has the same affect as DF_1_CONFALT.
Initially, the global offset table holds information as required by its relocation entries.
After the system creates memory segments for a loadable object file, the runtime linker
processes the relocation entries, some of which will be type R_SPARC_GLOB_DAT (for
SPARC), or R_386_GLOB_DAT (for x86), referring to the global offset table.
The runtime linker determines the associated symbol values, calculates their absolute
addresses, and sets the appropriate memory table entries to the proper values.
Although the absolute addresses are unknown when the link-editor creates an object
file, the runtime linker knows the addresses of all memory segments and can thus
calculate the absolute addresses of the symbols contained therein.
The tables entry zero is reserved to hold the address of the dynamic structure,
referenced with the symbol _DYNAMIC. This symbol enables a program, such as the
runtime linker, to find its own dynamic structure without having yet processed its
relocation entries. This method is especially important for the runtime linker, because
it must initialize itself without relying on other programs to relocate its memory
image.
The system can choose different memory segment addresses for the same shared
object in different programs. It can even choose different library addresses for different
executions of the same program. Nonetheless, memory segments do not change
addresses once the process image is established. As long as a process exists, its
memory segments reside at fixed virtual addresses.
A global offset tables format and interpretation are processor-specific. For SPARC and
x86 processors, the symbol _GLOBAL_OFFSET_TABLE_ can be used to access the
table. This symbol can reside in the middle of the .got section, allowing both
negative and nonnegative subscripts into the array of addresses. The symbol type is an
array of Elf32_Addr for 32bit code, and an array of Elf64_Addr for 64bit code:
extern Elf32_Addr _GLOBAL_OFFSET_TABLE_[];
extern Elf64_Addr _GLOBAL_OFFSET_TABLE_[];
A relocation table is associated with the procedure linkage table. The DT_JMP_REL
entry in the _DYNAMIC array gives the location of the first relocation entry. The
relocation table has one entry, in the same sequence, for each non-reserved procedure
linkage table entry. The relocation type of each of these entries is
R_SPARC_JMP_SLOT. The relocation offset specifies the address of the first byte of the
associated procedure linkage table entry. The symbol table index refers to the
appropriate symbol.
To illustrate procedure linkage tables, Table 747 shows four entries: two of the four
initial reserved entries, the third is a call to name101, and the fourth entry is a call to
name102. The example assumes that the entry for name102 is the tables last entry
and shows the following nop instruction. The left column shows the instructions from
the object file before dynamic linking. The right column demonstrates a possible way
the runtime linker might fix the procedure linkage table entries.
.PLT0: .PLT0:
unimp save %sp, -64, %sp
unimp call runtime_linker
unimp nop
.PLT1: .PLT1:
unimp .word identification
unimp unimp
unimp unimp
.PLT101: .PLT101:
sethi (.-.PLT0), %g1 nop
ba,a .PLT0 ba,a name101
nop nop
.PLT102: .PLT102:
sethi (.-.PLT0), %g1 sethi (.-.PLT0), %g1
ba,a .PLT0 sethi %hi(name102), %g1
nop jmpl %g1+%lo(name102), %g0
nop nop
Following the steps below, the runtime linker and program jointly resolve the
symbolic references through the procedure linkage table. Again, the steps described
below are for explanation only. The precise execution-time behavior of the runtime
linker is not specified.
The runtime linker does not have to create the instruction sequences under the
memory segment column. If it does, some points deserve more explanation.
To make the code re-entrant, the procedure linkage tables instructions are changed
in a particular sequence. If the runtime linker is fixing a functions procedure
linkage table entry and a signal arrives, the signal handling code must be able to
call the original function with predictable and correct results.
The runtime linker changes three words to convert an entry. The runtime linker can
update only a single word atomically with regard to instruction execution.
Therefore, re-entrancy is achieved by updating each word in reverse order. If a
re-entrant function call occurs just prior to the last patch, the runtime linker gains
control a second time. Although both invocations of the runtime linker modify the
same procedure linkage table entry, their changes do not interfere with each other.
The first sethi instruction of a procedure linkage table entry can fill the delay slot
of the previous entrys jmp1 instruction. Although the sethi changes the value of
the %g1 register, the previous contents can be safely discarded.
After conversion, the last procedure linkage table entry, .PLT102, needs a delay
instruction for its jmp1. The required, trailing nop fills this delay slot.
Note The different instruction sequences shown for .PLT101, and .PLT102
demonstrate how the update may be optimized for the associated destination.
The first four procedure linkage table entries are reserved. The original contents of
these entries are unspecified, despite the example shown in Table 748. Each of the
first 32,768 entries in the table occupies 8 words (32 bytes), and must be aligned on a
32byte boundary. The table as a whole must be aligned on a 256byte boundary. If
more than 32,768 entries are required, the remaining entries consist of 6 words (24
bytes) and 1 pointer (8 bytes). The instructions are collected together in blocks of 160
entries followed by 160 pointers. The last group of entries and pointers may contain
less than 160 items. No padding is required.
Note The numbers 32,768 and 160 are based on the limits of branch and load
displacements respectively with the second rounded down to make the divisions
between code and data fall on 256byte boundaries so as to improve cache
performance.
A relocation table is associated with the procedure linkage table. The DT_JMP_REL
entry in the _DYNAMIC array gives the location of the first relocation entry. The
relocation table has one entry, in the same sequence, for each non-reserved procedure
linkage table entry. The relocation type of each of these entries is
R_SPARC_JMP_SLOT. For the first 32,767 slots, the relocation offset specifies the
address of the first byte of the associated procedure linkage table entry, the addend
field is zero. The symbol table index refers to the appropriate symbol. For slots 32,768
and beyond, the relocation offset specifies the address of the first byte of the associated
pointer. The addend field is the unrelocated value -(.PLTN + 4). The symbol table
index refers to the appropriate symbol.
To illustrate procedure linkage tables, Table 748 shows several entries. The first three
show initial reserved entries. The following three show examples of the initial 32,768
entries together with possible resolved forms that might apply if the target address
was +/- 2 Gbytes of the entry, within the lower 4 Gbytes of the address space, or
anywhere respectively. The final two show examples of later entries, which consist of
instruction and pointer pairs. The left column shows the instructions from the object
file before dynamic linking. The right column demonstrates a possible way the
runtime linker might fix the procedure linkage table entries.
.PLT0: .PLT0:
unimp save %sp, -176, %sp
unimp sethi %hh(runtime_linker_0), %l0
unimp sethi %lm(runtime_linker_0), %l1
unimp or %l0, %hm(runtime_linker_0), %l0
unimp sllx %l0, 32, %l0
unimp or %l0, %l1, %l0
unimp jmpl %l0+%lo(runtime_linker_0), %o1
unimp mov %g1, %o0
.PLT1: .PLT1:
unimp save %sp, -176, %sp
unimp sethi %hh(runtime_linker_1), %l0
unimp sethi %lm(runtime_linker_1), %l1
unimp or %l0, %hm(runtime_linker_1), %l0
unimp sllx %l0, 32, %l0
unimp or %l0, %l1, %l0
unimp jmpl %l0+%lo(runtime_linker_0), %o1
unimp mov %g1, %o0
.PLT2: .PLT2:
unimp .xword identification
.PLT101: .PLT101:
sethi (.-.PLT0), %g1 nop
ba,a %xcc, .PLT1 mov %o7, %g1
nop call name101
nop mov %g1, %o7
nop; nop nop; nop
nop; nop nop; nop
.PLT102: .PLT102:
sethi (.-.PLT0), %g1 nop
ba,a %xcc, .PLT1 sethi %hi(name102), %g1
nop jmpl %g1+%lo(name102), %g0
nop nop
nop; nop nop; nop
nop; nop nop; nop
.PLT103: .PLT103:
sethi (.-.PLT0), %g1 nop
ba,a %xcc, .PLT1 sethi %hh(name103), %g1
nop sethi %lm(name103), %g5
nop or %hm(name103), %g1
nop sllx %g1, 32, %g1
nop or %g1, %g5, %g5
nop jmpl %g5+%lo(name103), %g0
nop nop
.PLT32768: .PLT32768:
mov %o7, %g5 <unchanged>
call .+8 <unchanged>
nop <unchanged>
ldx [%o7+.PLTP32768 - <unchanged>
(.PLT32768+4)], %g1
jmpl %o7+%g1, %g1 <unchanged>
mov %g5, %o7 <unchanged>
... ...
.PLT32927: .PLT32927:
mov %o7, %g5 <unchanged>
call .+8 <unchanged>
nop <unchanged>
ldx [%o7+.PLTP32927 - <unchanged>
(.PLT32927+4)], %g1
jmpl %o7+%g1, %g1 <unchanged>
mov %g5, %o7 <unchanged>
.PLTP32768 .PLTP32768
.xword .PLT0 - .xword name32768 -
(.PLT32768+4) (.PLT32768+4)
... ...
.PLTP32927 .PLTP32927
.xword .PLT0 - .xword name32927 -
(.PLT32927+4) (.PLT32927+4)
Following the steps below, the runtime linker and program jointly resolve the
symbolic references through the procedure linkage table. Again, the steps described
below are for explanation only. The precise execution-time behavior of the runtime
linker is not specified.
1. When first creating the memory image of the program, the runtime linker changes
the initial procedure linkage table entries, making them transfer control to one of
the runtime linkers own routines. The runtime linker also stores an extended word
of identification information in the third entry. When the runtime linker receives
control, it can examine this extended word to find which object called it.
2. All other procedure linkage table entries initially transfer to the first or second
entry. Those entries establish a stack frame and call the runtime linker.
3. With the identification value, the runtime linker gets its data structures for the
object, including the relocation table.
4. The runtime linker computes the index of the relocation entry for the table slot.
The runtime linker does not have to create the instruction sequences under the
memory segment column, it might. If it does, some points deserve more explanation.
To make the code re-entrant, the procedure linkage tables instructions are changed
in a particular sequence. If the runtime linker is fixing a functions procedure
linkage table entry and a signal arrives, the signal handling code must be able to
call the original function with predictable and correct results.
The runtime linker may change up to eight words to convert an entry. The runtime
linker can update only a single word atomically with regard to instruction
execution. Therefore, re-entrancy is achieved by first overwriting the nop
instructions with their replacement instructions, and then patching the ba,a, and
the sethi if using a 64bit store. If a re-entrant function call occurs just prior to the
last patch, the runtime linker gains control a second time. Although both
invocations of the runtime linker modify the same procedure linkage table entry,
their changes do not interfere with each other.
If the initial sethi instruction is changed, it can only be replaced by a nop.
Changing the pointer as done for the second form of entry is done using a single
atomic 64bit store.
Note The different instruction sequences shown for .PLT101, .PLT102, and
.PLT103 demonstrate how the update may be optimized for the associated
destination.
.PLT0:
pushl got_plus_4
jmp *got_plus_8
nop; nop
nop; nop
.PLT1:
jmp *name1_in_GOT
pushl $offset
jmp .PLT0@PC
.PLT2:
jmp *name2_in_GOT
pushl $offset
jmp .PLT0@PC
.PLT0:
pushl 4(%ebx)
jmp *8(%ebx)
nop; nop
nop; nop
.PLT1:
jmp *name1@GOT(%ebx)
pushl $offset
jmp .PLT0@PC
.PLT2:
jmp *name2@GOT(%ebx)
pushl $offset
jmp .PLT0@PC
Note As the preceding examples show, the procedure linkage table instructions use
different operand addressing modes for absolute code and for position-independent
code. Nonetheless, their interfaces to the runtime linker are the same.
Following the steps below, the runtime linker and program cooperate to resolve the
symbolic references through the procedure linkage table and the global offset table.
1. When first creating the memory image of the program, the runtime linker sets the
second and third entries in the global offset table to special values. The steps below
explain these values.
2. If the procedure linkage table is position-independent, the address of the global
offset table must be in %ebx. Each shared object file in the process image has its
own procedure linkage table, and control transfers to a procedure linkage table
entry only from within the same object file. So, the calling function must set the
global offset table base register before it calls the procedure linkage table entry.
Thread-Local Storage
The compilation environment supports the declaration of thread-local data. This data
is sometime referred to as thread-specific, or thread-private data, but more typically by
the acronym TLS. By declaring variables to be thread-local, the compiler automatically
arranges for these variables to be allocated on a per-thread basis.
271
Initialization
In C++, a thread-local variable may not be initialized if the initialization requires a
static constructor. Otherwise, a thread-local variable may be initialized to any value
that would be legal for an ordinary static variable.
Static TLS models generates faster code. However, code compiled to use this model
cannot reference thread-local variables in post-startup dynamically loaded libraries.
A dynamic TLS model is able to reference all TLS. These models are described in
Thread-Local Storage Access Models on page 277.
Address-of operator
The address-of operator, &, can be applied to a thread-local variable. This operator
is evaluated at runtime, and returns the address of the variable within the current
thread. The address obtained by this operator may be used freely by any thread in
the process as long as the thread that evaluated the address remains in existence.
When a thread terminates, any pointers to thread-local variables in that thread
become invalid.
The compilation environment allocates TLS in sections identified with the SHF_TLS
flag. These sections provide initialized and uninitialized TLS based on how the storage
is declared:
An initialized thread-local variable is allocated in a .tdata, or .tdata1 section.
This initialization may require relocation.
The initialized portion of this template is called the TLS initialization image. All
relocations generated as a result of initialized thread-local variables are applied to this
template. These relocated values are then used when a new thread requires the initial
values.
TLS symbols have the symbol type STT_TLS. These symbols are assigned offsets
relative to the beginning of the TLS template. The actual virtual address associated
with these symbols is irrelevant. The address refers only to the template, and not to
the per-thread copy of each data item.
In dynamic executables and shared objects, the st_value field of a STT_TLS symbol
contains the assigned offset for defined symbols, or zero for undefined symbols.
Several relocations are defined to support access to TLS. See SPARC: Thread-Local
Storage Relocation Types on page 283 and x86: Thread-Local Storage Relocation
Types on page 290. Symbols of type STT_TLS are only referenced by TLS relocations.
TLS relocations only reference symbols of type STT_TLS.
In dynamic executables and shared objects, a PT_TLS program entry describes a TLS
template, and has the following members:
Member Value
p_paddr Reserved
p_flags PF_R
dtvt
gent dtvt,1 dtvt,2 dtvt,3 dtvt,4 dtvt,5
Program Startup
At program startup, the runtime system creates TLS for the main thread.
First, the runtime linker logically combines the TLS templates for all loaded dynamic
objects, including the dynamic executable, into a single static template. Each dynamic
objectss TLS template is assigned an offset within the combined template,
tlsoffsetm, as follows:
tlsoffset1 = round(tlssize1, align1)
tlsoffsetm+1 = round(tlsoffsetm + tlssizem+1, alignm+1)
tlssizem+1 and alignm+1 are the size and alignment, respectively, for the allocation
template for dynamic object m (1 <= m <= M, where M is the total number of loaded
dynamic objects). The round(offset, align) function returns an offset rounded
up to the next multiple of align. The TLS template is placed immediately preceding the
thread pointer tpt. Accesses to the TLS data are based off of subtractions from tpt.
The runtime linker then constructs a linked list of initialization records. Each record in
this list describes the TLS initialization image for one loaded dynamic object, and
contains the following fields:
A pointer to the TLS initialization image.
The size of the TLS initialization image.
The tlsoffsetm of the object.
A flag indicating whether the object uses a static TLS model.
The thread library uses this information to allocate storage for the initial thread. This
storage is initialized, and a dynamic TLS vector for the initial thread is created.
Thread Creation
For the initial thread, and for each new thread created, the thread library allocates a
new TLS block for each loaded dynamic object. Blocks may be allocated separately, or
as a single contiguous block.
Each thread t has an associated thread pointer tpt, which points to the thread control
block, TCB. The thread pointer, tp, always contains the value of tpt for the current
running thread.
The thread library then creates a vector of pointers, dtvt, for the current thread t. The
first element of each vector contains a generation number gent, which is used to
determine when the vector needs to be extended. See Deferred Allocation of
Thread-Local Storage Blocks on page 276.
Each remaining element in the vector, dtvt,m, is a pointer to the block reserved for the
TLS belonging to dynamic object m.
For dynamically loaded, post-startup objects, the thread library defers the allocation of
TLS blocks. This allocation occurs when the first reference is made to a TLS variable
within the loaded object. All references to TLS defined in a post-startup, dynamically
loaded object, must use a dynamic TLS model. For blocks whose allocation has been
deferred, the pointer dtvt,m is set to an implementation-defined special value.
Note The runtime linker may group TLS templates for all startup objects such that
they share a single element in the vector, dtvt,1. This does not affect the offset
calculations described above or the creation of the list of initialization records. For the
following sections, however, the value of M, the total number of objects, start with the
value of 1.
When a library containing TLS is unloaded, the TLS blocks used by that library are
freed.
Note The SPARC and x86 definitions of this function have the same function
signature. However, the x86 version does not use the default x86 calling convention of
passing arguments on the stack. Instead, the x86 version passes its argument via the
%eax register which is more efficient. To denote that this alternate calling method is
used, the x86 function name has three leading underscores in its name.
The link-editor can transition code from the more general access models to the more
optimized models, if it is determined appropriate to do so. This transitioning is
achievable through the use of unique TLS relocations. These relocations, not only
request updates be performed, but identify which TLS access model is being used.
Knowing the TLS access model, and the type of object being created, allows the
link-editor to perform translations. For example, if a relocatable object using the GD
access model is being linked into a dynamic executable, the link-editor can transition
the references using the IE or LE access models, as appropriate. The relocations
required for the model are then performed.
The following diagram illustrates the different access models, and when one model
can be transitioned from one to the other.
Backend
command-line
General Intitial
dynamic exec
Backend known
local optimization
Linker known
exec optimization
Linker known
local optimization
Default
Optimization
GOT[n] R_SPARC_TLS_DTPMOD32 x
GOT[n + 1] R_SPARC_TLS_DTPOFF32 x
GOT[n] R_SPARC_TLS_DTPMOD64 x
GOT[n + 1] R_SPARC_TLS_DTPOFF64 x
Since the load object index and TLS block index for x are not known until runtime, the
link-editor places the R_SPARC_TLS_DTPMOD32 and R_SPARC_TLS_DPTOFF32
relocations against the GOT for processing by the runtime linker.
The register used as the GOT-pointer for the add instruction tagged by the
R_SPARC_TLS_GD_ADD relocation, must be the first register in the add instruction.
This permits the link-editor to identify the GOT-pointer register during a code
transformation.
TABLE 83 SPARC: 32-bit and 64-bit Local Dynamic Thread-Local Variable Access Codes
GOT[n] R_SPARC_TLS_DTPMOD32 x1
GOT[n + 1] <none>
GOT[n] R_SPARC_TLS_DTPMOD64 x1
GOT[n + 1] <none>
Since the load object index is not known until runtime, a R_SPARC_TLS_DTPMOD32
relocation is created, and the ti_tlsoffset field of the TLS_index structure is zero
filled.
The second add and the call instruction are tagged with the
R_SPARC_TLS_LDM_ADD and R_SPARC_TLS_LDM_CALL relocations respectively.
When a procedure references more then one local symbol, the compiler generates code
to obtain the base address of the TLS block once. This base address is then used to
calculate the address of each symbol without a separate library call.
Note The register containing the TLS object address in the add instruction tagged by
the R_SPARC_TLS_LDO_ADD must be the first register in the instruction sequence.
This permits the link-editor to identify the register during a code transformation.
GOT[n] R_SPARC_TLS_TPOFF32 x
Note The register used as the GOT-pointer for the add instruction tagged by the
R_SPARC_TLS_IE_ADD relocation must be the first register in the instruction. This
permits the link-editor to identify the GOT-pointer register during a code
transformation.
GOT[n] R_SPARC_TLS_TPOFF64 x
TABLE 86 SPARC: 32-bit and 64-bit Local Executable Thread-Local Variable Access Codes
GOT[n] R_386_TLS_DTPMOD32 x
GOT[n + 1] R_386_TLS_DTPOFF32
Since the load object index and TLS block index for x are not known until runtime, the
link-editor places the R_386_TLS_DTPMOD32 and R_386_TLS_DTPOFF32 relocations
against the GOT for processing by the runtime linker. The address of the generated GOT
entry is loaded into register %eax for the call to ___tls_get_addr().
The call instruction must immediately follow the leal instruction. This is a required
to permit the code transformations.
GOT[n] R_386_TLS_DTPMOD32 x
GOT[n + 1] <none>
The first leal instruction generates a R_386_TLS_LDM relocation that instructs the
link-editor to allocate space in the global offset table to hold a TLS_index structure
for the current object. The link-editor process this relocation by substituting the
GOT-relative offset for the new linkage table entry.
Since the load object index is not known until runtime, aR_386_TLS_DTPMOD32
relocation is created, and the ti_tlsoffset field of the structure is zero filled. The
call instruction is tagged with the R_386_TLS_LDM_PLT relocation.
The TLS offset for each local symbol is known at link-edit time so the link-editor fills
these values in directly.
When a procedure references more then one local symbol, the compiler generates code
to obtain the base address of the TLS block once. This base address is then used to
calculate the address of each symbol without a separate library call.
TABLE 810 x86: Initial Executable, Position Independent, Thread-Local Variable Access
Codes
GOT[n] R_386_TLS_TPOFF x
TABLE 811 x86: Initial Executable, Position Dependent, Thread-Local Variable Access Codes
GOT[n] R_386_TLS_TPOFF x
The contents of variable x, rather then the address, can be loaded by embedding the
offset directly into the memory reference as shown in the next two sequences.
TABLE 812 x86: Initial Executable, Position Independent, Dynamic Thread-Local Variable
Access Codes
GOT[n] R_386_TLS_TPOFF x
TABLE 813 x86: Initial Executable, Position Independent, Thread-Local Variable Access
Codes
GOT[n] R_386_TLS_TPOFF x
In the last sequence, if the %eax register is used instead of the %ecx above, the first
instruction may be either 5 or 6 bytes long.
The contents of variable x, rather then the address, can be accessed with the same
relocation by using the following instruction sequence.
If instead of computing the address of the variable we want to load from it or store in
it the following sequence can be used. Note that in this case we use the x@ntpoff
expression not as an immediate value, but instead as an absolute address.
Mapfile Option
The link-editor automatically and intelligently maps input sections from relocatable
objects to segments in the output file being created. The -M option with an associated
mapfile enables you to change the default mapping provided by the link-editor. In
addition, new segments can be created, attributes modified, and symbol versioning
information can be supplied with the mapfile.
Note When using a mapfile option, you can easily create an output file that does
not execute. The link-editor knows how to produce a correct output file without the
use of the mapfile option.
Each directive can span more than one line and can have any amount of white space,
including new lines, as long as that white space is followed by a semicolon.
293
Typically, segment declarations are followed by mapping directives. You declare a
segment and then define the criteria by which a section becomes part of that segment.
If you enter a mapping directive or size-symbol declaration without first declaring the
segment to which you are mapping, except for built-in segments, the segment is given
default attributes. Such segment is an implicitly declared segment.
The following sections describe each directive type. For all syntax discussions, the
following notations apply:
All entries in constant width, all colons, semicolons, equal signs, and at (@) signs
are typed in literally.
All entries in italics are substitutable.
{ ... }* means zero or more.
{ ... }+ means one or more.
[ ... ] means optional.
section_names and segment_names follow the same rules as C identifiers,
where a period (.) is treated as a letter. For example, .bss is a legal name.
section_names, segment_names, file_names, and symbol_names are case
sensitive. Everything else is not case sensitive.
Spaces, or new-lines, can appear anywhere except before a number or in the
middle of a name or value.
Comments beginning with # and ending at a newline can appear anywhere that a
space can appear.
Segment Declarations
A segment declaration creates a new segment in the output file, or changes the
attribute values of an existing segment. An existing segment is one that you previously
defined or one of the four built-in segments described immediately following.
Attribute Value
virtual_address Vnumber
physical_address Pnumber
length Lnumber
rounding Rnumber
alignment Anumber
There are four built-in segments with the following default attribute values:
text LOAD, ?RX, no virtual_address, physical_address, or length
specified, alignment values set to defaults per CPU type.
data LOAD, ?RWX, no virtual_address, physical_address, or length
specified, alignment values set to defaults per CPU type.
bss disabled, LOAD, ?RWX, no virtual_address, physical_address, or
length specified, alignment values set to defaults per CPU type.
note NOTE.
By default, the bss segment is disabled. Any sections of type SHT_NOBITS, which are
its sole input, are captured in the data segment. See Table 712 for a full description
of SHT_NOBITS sections. The simplest bss declaration:
bss =;
The link-editor behaves as if these segments are declared before your mapfile is read
in. See Mapfile Option Defaults on page 302.
Note The link-editor calculates the addresses and length of the current segment
based on the previous segments attribute values.
The ?E flag allows the creation of an empty segment. This empty segment has no
sections associated with it. This segment can only be specified for executables, and
must be of type LOAD with a specified size and alignment. Multiple segment
definitions of this type are permitted.
The ?N flag enables you control whether the ELF header, and any program headers are
included as part of the first loadable segment. By default, the ELF header and program
headers are included with the first segment. The information in these headers is used
within the mapped image, typically by the runtime linker. The use of the ?N option
causes the virtual address calculations for the image to start at the first section of the
first segment.
The ?O flag enables you control the order of sections in the output file. This flag is
intended for use in conjunction with the -xF option to the compilers. When a file is
compiled with the -xF option, each function in that file is placed in a separate section
with the same attributes as the .text section. These sections are called
.text%function_name.
For example, a file containing three functions, main(), foo() and bar(), when
compiled with the -xF option, yields a relocatable object file with text for the three
functions being placed in sections called .text%main, .text%foo, and .text%bar.
Because the -xF option forces one function per section, the use of the ?O flag to control
the order of sections in effect controls the order of functions.
The first declaration associates the ?O flag with the default text segment.
If the order of function definitions in the source file is main, foo, and bar, then the
final executable contains functions in the order foo, bar, and main.
For static functions with the same name, the file names must also be used. The ?O flag
forces the ordering of sections as requested in the mapfile. For example, if a static
function bar() exists in files a.o and b.o, and function bar() from file a.o is to be
placed before function bar() from file b.o, then the mapfile entries should read:
text: .text%bar: a.o;
text: .text%bar: b.o;
this entry does not guarantee that function bar() from file a.o is placed before
function bar() from file b.o. The second format is not recommended as the results
are not reliable.
Mapping Directives
A mapping directive instructs the link-editor how to map input sections to output
segments. Basically, you name the segment that you are mapping to and indicate what
the attributes of a section must be in order to map into the named segment. The set of
section_attribute_values that a section must have to map into a specific
segment is called the entrance criteria for that segment. In order to be placed in a
specified segment of the output file, a section must meet the entrance criteria for a
segment exactly.
section_type $PROGBITS
$SYMTAB
$STRTAB
$REL
$RELA
$NOTE
$NOBITS
Entering more than one mapping directive line for a segment is the only way to
specify multiple values of a section attribute.
A section can match more than one entrance criteria. In this case, the first segment
encountered in the mapfile with that entrance criteria is used. For example, if a
mapfile reads:
S1 : $PROGBITS;
S2 : $PROGBITS;
Section-Within-Segment Ordering
By using the following notation you can specify the order that sections are placed
within a segment:
segment_name | section_name1;
segment_name | section_name2;
segment_name | section_name3;
The sections that are named in the above form are placed before any unnamed
sections, and in the order they are listed in the mapfile.
symbol_name can be any legal C identifier. The link-editor does not check the syntax
of the symbol_name.
Mapping Example
The following example is a user-defined mapfile. The numbers on the left are
included in the example for tutorial purposes. Only the information to the right of the
numbers actually appears in the mapfile.
Four separate segments are manipulated in this example. The implicitly declared
segment elephant (line 1) receives all of the .data sections from the files
peanuts.o and popcorn.o. Notice that *popcorn.o matches any popcorn.o file
that can be supplied to the link-edit. The file need not be in the current directory. On
the other hand, if /var/tmp/peanuts.o was supplied to the link-edit, it does not
match peanuts.o because it is not preceded by an *.
The monkey segment is implicitly declared in line 2 with segment_type value LOAD,
segment_flags value RWX, and no virtual_address, physical_address,
length or alignment values specified (defaults are used). In line 4 the
segment_type value of monkey is set to LOAD. Because the segment_type attribute
value does not change, no warning is issued. The virtual_address value is set to
0x80000000 and the maximum length value to 0x4000.
Line 5 implicitly declares the donkey segment. The entrance criteria are designed to
route all .data sections to this segment. Actually, no sections fall into this segment
because the entrance criteria for monkey in line 3 capture all of these sections. In line
6, the segment_flags value is set to ?RX and the alignment value is set to
0x1000. Because both of these attribute values changed, a warning is issued.
The text and data segments are manipulated in this example. Line 1 declares the
text segment to have a virtual_address of 0xf0004000 and to not include the
ELF header or any program headers as part of this segments address calculations.
Any other $PROGBITS section that makes up the text segment follows the .rodata
section. Line 5 declares the data segment and specifies that its virtual address must
begin on a 0x1000 byte boundary. The first section that constitutes the data segment
also resides on a 0x1000 byte boundary within the file image.
The following example shows how a mapfile would appear for the link-editor
defaults. The link-editor begins execution behaving as if the mapfile has already
been read in. Then the link-editor reads your mapfile and either augments or makes
changes to the defaults.
text = LOAD ?RX;
text : ?A!W;
data = LOAD ?RWX;
data : ?AW;
note = NOTE;
note : $NOTE;
As each segment declaration in your mapfile is read in, it is compared to the existing
list of segment declarations as follows:
1. If the segment does not already exist in the mapfile but another with the same
segment-type value exists, the segment is added before all of the existing segments
of the same segment_type.
2. If none of the segments in the existing mapfile has the same segment_type
value as the segment just read in, then the segment is added by segment_type
value to maintain the following order:
INTERP
LOAD
DYNAMIC
NOTE
As each mapping directive in a mapfile is read in, the directive is added after any
other mapping directives that you already specified for the same segment but before
the default mapping directives for that segment.
A typical although somewhat simplified map structure is illustrated in Figure 91. The
Entrance Criteria boxes correspond to the information in the default mapping
directives. The Segment Attribute Descriptors boxes correspond to the information
in the default segment declarations. The Output Section Descriptors boxes give the
detailed attributes of the sections that fall under each segment. The sections
themselves are shown in circles.
.data .data
$PROGBITS from
?AWX fido.o
.bss .bss
$NOBITS from
?AWX rover.o
The link-editor performs the following steps when mapping sections to segments:
1. When a section is read in, the link-editor checks the list of Entrance Criteria looking
for a match. All specified criteria must be matched.
In Figure 91, a section that falls into the text segment must have a
section_type value of $PROGBITS and have a section_flags value of ?A!W.
It need not have the name .text since no name is specified in the Entrance
Criteria. The section can be either X or !X in the section_flags value because
nothing was specified for the execute bit in the Entrance Criteria.
If no Entrance Criteria match is found, the section is placed at the end of the output
file after all other segments. No program header entry is created for this
information.
2. When the section falls into a segment, the link-editor checks the list of existing
Output Section Descriptors in that segment as follows:
If the section attribute values match those of an existing Output Section Descriptor
exactly, the section is placed at the end of the list of sections associated with that
Output Section Descriptor.
Note If the input section has a user-defined section type value between
SHT_LOUSER and SHT_HIUSER, it is treated as a $PROGBITS section. No method
exists for naming this section_type value in the mapfile, but these sections
can be redirected using the other attribute value specifications (section_flags,
section_name) in the entrance criteria.
3. If a segment contains no sections after all of the command line object files and
libraries are read in, no program header entry is produced for that segment.
Note Input sections of type $SYMTAB, $STRTAB, $REL, and $RELA are used
internally by the link-editor. Directives that refer to these section types can only map
output sections produced by the link-editor to segments.
The following sections provide a simple overview, or cheat sheet, of the most
commonly used link-editor scenarios. See Link-Editing on page 20 for an
introduction to the kinds of output modules generated by the link-editor.
The examples provided show the link-editor options as supplied to a compiler driver,
this being the most common mechanism of invoking the link-editor. In these examples
we use cc(1). See Using a Compiler Driver on page 27.
The link-editor places no meaning on the name of any input file. Each file is opened
and inspected to determine the type of processing it requires. See Input File
Processing on page 28.
Shared objects that follow a naming convention of libx.so, and archive libraries that
follow a naming convention of libx.a, can be input using the -l option. See Library
Naming Conventions on page 31. This provides additional flexibility in allowing
search paths to be specified using the -L option. See Directories Searched by the
Link-Editor on page 33.
Static Mode
Static mode is selected when the -d n option is used, and enables you to create
relocatable objects and static executables. Under this mode, only relocatable objects
and archive libraries are acceptable forms of input. Use of the -l option results in a
search for archive libraries.
307
Creating a Relocatable Object
To create a relocatable object use the -d n and -r options:
$ cc -dn -r -o temp.o file1.o file2.o file3.o .....
The -a option is available to indicate the creation of a static executable. The use of
-d n without a -r implies -a.
Dynamic Mode
Dynamic mode is the default mode of operation for the link-editor. It can be enforced
by specifying the -d y option, but is implied when not using the -d n option.
Under this mode, relocatable objects, shared objects and archive libraries are
acceptable forms of input. Use of the -l option results in a directory search, where
each directory is searched for a shared object. If no shared object is found, the same
directory is then searched for an archive library. A search only for archive libraries can
be enforced by using the -B static option. See Linking With a Mix of Shared
Objects and Archives on page 32.
If the shared object being generated is used as input to another link-edit, record
within it the shared objects runtime name using the -h option. See Recording a
Shared Object Name on page 103.
Make the shared object available to the compilation environment by creating a file
system link to a non-versioned shared object name. See Coordination of Versioned
Filenames on page 146.
ELF objects make available global symbols to which other objects can bind. Some of
these global symbols can be identified as providing the objects public interface. Other
symbols are part of the objects internal implementation and are not intended for
external use. An objects interface can evolve from one software release to another. The
ability to identify this evolution is desirable.
Shared objects are prime candidates for internal versioning. This technique defines
their evolution, provides for interface validation during runtime processing (see
Binding to a Version Definition on page 136), and provides for the selective binding
of applications (see Specifying a Version Binding on page 140). Shared objects are
used as the examples throughout this appendix.
The following sections provide a simple overview, or cheat sheet, of the internal
versioning mechanism provided by the link-editors as applied to shared objects. The
examples recommend conventions and mechanisms for versioning shared objects,
from their initial construction through several common update scenarios.
Naming Conventions
A shared object follows a naming convention that includes a major number file suffix.
See Naming Conventions on page 102. Within this shared object, one or more version
definitions can be created. Each version definition corresponds to one of the following
categories:
311
It defines an industry-standard interface (for example, the System V Application
Binary Interface).
It defines a vendor-specific public interface.
It defines a vendor-specific private interface.
It defines a vendor-specific change to the internal implementation of the object.
The following version definition naming conventions help indicate which of these
categories the definition represents.
The first three of these categories indicate interface definitions. These definitions
consist of an association of the global symbol names that make up the interface, with a
version definition name. See Creating a Version Definition on page 131. Interface
changes within a shared object are often referred to as minor revisions. Therefore,
version definitions of this type are suffixed with a minor version number, which is
based on the file names major version number suffix.
The last category indicates a change having occurred within the object. This definition
consists of a version definition acting as a label and has no symbol name associated
with it. This definition is referred to as being a weak version definition. See Creating
a Weak Version Definition on page 134. Implementation changes within a shared
object are often referred to as micro revisions. Therefore, version definitions of this
type are suffixed with a micro version number based on the previous minor number to
which the internal changes have been applied.
Any industry standard interface should use a version definition name that reflects the
standard. Any vendor interfaces should use a version definition name unique to that
vendor. The companys stock symbol is often appropriate.
Private version definitions indicate symbols that have restricted or uncommitted use,
and should have the word private clearly visible.
All version definitions result in the creation of associated version symbol names. The
use of unique names and the minor/micro suffix convention reduces the chance of
symbol collision within the object being built.
The following version definition examples show the possible use of these naming
conventions:
SVABI.1
Defines the System V Application Binary Interface standards interface.
SUNW_1.1
Defines a Solaris public interface.
SUNWprivate_1.1
Defines a Solaris private interface.
SUNW_1.1.1
Defines a Solaris internal implementation change.
Vendor private interfaces are very unstable, and can change, or even be deleted, from
release to release. These interfaces provide for uncommitted or experimental
functionality, or are intended to provide access for vendor-specific applications only. If
you want to achieve any level of binary compatibility, you should avoid using these
interfaces.
Any global symbols that do not fall into one of the above categories should be reduced
to local scope so that they are no longer visible for binding. See Reducing Symbol
Scope on page 52.
The global symbols foo1 and foo2 are assigned to the shared objects public interface
SUNW_1.1. Any other global symbols supplied from the input files are reduced to
local by the auto-reduction directive *. See Reducing Symbol Scope on page 52.
Compatible updates that can be accommodated by internal versioning fall into three
basic categories:
Adding new symbols
Creating new interfaces from existing symbols
Internal implementation changes
The first two categories are achieved by associating an interface version definition
with the appropriate symbols. The latter is achieved by creating a weak version
definition that has no associated symbols.
The following mapfile example assigns the new symbol foo3 to the new interface
version definition SUNW_1.2. This new interface inherits the original interface
SUNW_1.1.
$ cat mapfile
SUNW_1.2 { # Release X+1.
global:
foo3;
} SUNW_1.1;
SUNW_1.1 { # Release X.
global:
foo2;
The inheritance of version definitions reduces the amount of version information that
must be recorded in any user of the shared object.
SUNW_1.1 { # Release X.
global:
foo2;
foo1;
local:
*;
};
SUNW_1.1 { # Release X.
Note The comments for the SUNW_1.1 and SUNW_1.1.1 version definitions indicate
that they have both been applied to the same release.
The following mapfile example shows the addition of a new industry standard
interface STAND.1. This interface contains the new symbol foo4 and the existing
symbols foo3 and foo1, which were originally offered through the interfaces
SUNW_1.2 and SUNW_1.1 respectively.
$ cat mapfile
STAND.1 { # Release X+2.
global:
foo4;
} STAND.0.1 STAND.0.2;
SUNW_1.1 { # Release X.
global:
foo2;
local:
*;
} STAND.0.2;
# Subversion - providing for
STAND.0.1 { # SUNW_1.2 and STAND.1 interfaces.
global:
foo3;
};
# Subversion - providing for
STAND.0.2 { # SUNW_1.1 and STAND.1 interfaces.
The symbols foo3 and foo1 are pulled into their own intermediate interface
definitions, which are used to create the original and new interface definitions.
The new definition of the SUNW_1.2 interface has referenced its own version
definition symbol. Without this reference, the SUNW_1.2 interface would have
contained no immediate symbol references and hence would be categorized as a weak
version definition.
Although the introduction of the new standards interface in software release X+2 has
changed the interface version definitions available, the list of symbols provided by
each of the original interfaces remains constant. The following example shows that
interface SUNW_1.2 still provides symbols foo1, foo2 and foo3.
$ pvs -ds -N SUNW_1.2 libfoo.so.1
SUNW_1.2:
STAND.0.1:
foo3;
SUNW_1.1:
foo2;
STAND.0.2:
foo1;
An application might only reference one of the new subversions. In this case, any
attempt to run the application on a previous release results in a runtime versioning
error. See Binding to a Version Definition on page 136.
main()
$ cat mapfile
libfoo.so - SUNW_1.1 $ADDVERS=SUNW_1.1;
$ cc -M mapfile -o prog prog.c -L. -R. -lfoo
$ pvs -r prog
libfoo.so.1 (SUNW_1.1);
In practice, you rarely have to promote a version binding in this manner. The
introduction of new standards binary interfaces is rare, and most applications
reference many symbols from an interface family.
A dynamic object can establish dependencies explicitly or through filters. Each of these
mechanisms can be augmented with a runpath, which directs the runtime linker to
search for and load the required dependency. String names used to record filters,
dependencies and runpath information can be augmented with the reserved dynamic
string tokens:
$ISALIST
$OSNAME, $OSREL and $PLATFORM
$ORIGIN
The following sections provide examples of how each of these tokens may be
employed.
Any string name that incorporates the $ISALIST token is effectively duplicated into
multiple strings. Each string is assigned one of the available instruction sets.
The following example shows how the auxiliary filter libfoo.so.1 can be designed
to access an instruction set specific filtee libbar.so.1.
$ LD_OPTIONS=-f /opt/ISV/lib/$ISALIST/libbar.so.1 \
cc -o libfoo.so.1 -G -K pic -h libfoo.so.1 -R. foo.c
$ dump -Lv libfoo.so.1 | egrep "SONAME|AUXILIARY"
[1] SONAME libfoo.so.1
[2] AUXILIARY /opt/ISV/lib/$ISALIST/libbar.so.1
321
$ LD_OPTIONS=-f libbar.so.1 \
cc -o libfoo.so.1 -G -K pic -h libfoo.so.1 -R/opt/ISV/lib/$ISALIST foo.c
$ dump -Lv libfoo.so.1 | egrep "RUNPATH|AUXILIARY"
[1] RUNPATH /opt/ISV/lib/$ISALIST
[2] AUXILIARY libbar.so.1
In either case the runtime linker uses the platform available instruction list to
construct multiple search paths. For example, the following application is dependent
on libfoo.so.1 and executed on a SUNW,Ultra-2:
$ ldd -ls prog
.....
find object=libbar.so.1; required by ./libfoo.so.1
search path=/opt/ISV/lib/$ISALIST (RPATH from file ./libfoo.so.1)
trying path=/opt/ISV/lib/sparcv9+vis/libbar.so.1
trying path=/opt/ISV/lib/sparcv9/libbar.so.1
trying path=/opt/ISV/lib/sparcv8plus+vis/libbar.so.1
trying path=/opt/ISV/lib/sparcv8plus/libbar.so.1
trying path=/opt/ISV/lib/sparcv8/libbar.so.1
trying path=/opt/ISV/lib/sparcv8-fsmuld/libbar.so.1
trying path=/opt/ISV/lib/sparcv7/libbar.so.1
trying path=/opt/ISV/lib/sparc/libbar.so.1
Any interface defined in a filter can result in an exhaustive search of all potential
filtees in an attempt to locate the required interface. If filtees are being employed to
provide performance critical functions, this exhaustive filtee searching can be
counterproductive.
A filtee can be built with the link-editors -z endfiltee option to indicate that it is
the last of the available filtees. This option terminates any further filtee searching for
that filter. From the previous SPARC example, if the sparcv9 filtee existed, and was
tagged with -z endfiltee, the filtee searches would be:
$OSNAME expands to reflect the name of the operating system, as displayed by the
utility uname(1) with the -s option. $OSREL expands to reflect the operating system
release level, as displayed by uname -r. $PLATFORM expands to reflect the
underlying hardware implementation, as displayed by uname -i.
The following example shows how the auxiliary filter libfoo.so.1 can be designed
to access a platform specific filtee libbar.so.1.
$ LD_OPTIONS=-f /usr/platform/$PLATFORM/lib/libbar.so.1 \
cc -o libfoo.so.1 -G -K pic -h libfoo.so.1 -R. foo.c
$ dump -Lv libfoo.so.1 | egrep "SONAME|AUXILIARY"
[1] SONAME libfoo.so.1
[2] AUXILIARY /usr/platform/$PLATFORM/lib/libbar.so.1
Assume that the product is designed for installation under /opt. Normally, you
would augment the PATH with /opt/ABC/bin to locate the products binaries. Each
binary locates their dependencies using a hard-coded runpath within the binary. For
the application abc, this runpath would be:
% cc -o abc abc.c -R/opt/ABC/lib -L/opt/ABC/lib -lA
% dump -Lv abc
[1] NEEDED libA.so.1
[2] RUNPATH /opt/ABC/lib
This dependency representation works until the product is installed in some directory
other than the recommended default.
The dynamic token $ORIGIN expands to the directory in which an object originated.
This token is available for filters, runpath, or dependency definitions. Use this
technology to redefine the unbundled application to locate its dependencies in terms
of $ORIGIN:
% cc -o abc abc.c -R$ORIGIN/../lib -L/opt/ABC/lib -lA
% dump -Lv abc
[1] NEEDED libA.so.1
[2] RUNPATH $ORIGIN/../lib
If this product is now installed under /usr/local/ABC and the users PATH is
augmented with /usr/local/ABC/bin, invocation of the application abc result in a
pathname lookup for its dependencies as follows:
% ldd -s abc
.....
For example, the unbundled product XYZ might have dependencies on the product
ABC. This dependency can be established by a host package installation script. This
script generates a symbolic link to the installation point of the ABC product, as shown
in the following figure.
XYZ
ABC
The binaries and shared objects of the XYZ product can represent their dependencies
on the ABC product using the symbolic link. This link is now a stable reference point.
For the application xyz, this runpath would be:
% cc -o xyz xyz.c -R$ORIGIN/../lib:$ORIGIN/../ABC/lib \
-L/opt/ABC/lib -lX -lA
% dump -Lv xyz
[1] NEEDED libX.so.1
[2] NEEDED libA.so.1
[3] RUNPATH $ORIGIN/../lib:$ORIGIN/../ABC/lib
If this product is now installed under /usr/local/XYZ, its post-install script would
be required to establish a symbolic link of:
% ln -s ../ABC /usr/local/XYZ/ABC
Security
In a secure process, the expansion of the $ORIGIN string is allowed only if it expands
to a trusted directory. The occurrence of other relative path names, poses a security
risk.
The following example shows this possible security breach if $ORIGIN was arbitrarily
expanded within a secure process.
% cd /worldwritable/dir/in/same/fs
% mkdir bin lib
% ln $ORIGIN/bin/program bin/program
% cp ~/crooked-libc.so.1 lib/libc.so.1
% bin/program
..... using crooked-libc.so.1
You can use the utility crle(1) to specify trusted directories that enable secure
applications to use $ORIGIN. Administrators who use this technique should ensure
that the target directories are suitably protected from malicious intrusion.
This appendix provides an overview of new features and updates that have been
added to the Solaris operating environment and indicates the release to which they
were added:
329
Solaris 9 8/03 Release
dlsym(3DL) symbol processing can be reduced using a dlopen(3DL) handle that
is created with the RTLD_FIRST flag. See Obtaining New Symbols on page 89.
The signal used by the runtime linker to terminate an erroneous process can be
managed using the dlinfo(3DL) flags RTLD_DI_GETSIGNAL, and
RTLD_DI_SETSIGNAL.
Solaris 9 Release
Thread-Local Storage (TLS) support is provided. See Chapter 8.
The -z rescan option provides greater flexibility in specifying archive libraries to
a link-edit. See Position of an Archive on the Command Line on page 32.
The -z ld32 and -z ld64 options provide greater flexibility in using the
link-editor support interfaces. See 32Bit and 64Bit Environments on page 150.
Additional link-editor support interfaces ld_input_done(),
ld_input_section(), ld_input_section64() and ld_version() have
been added. See Support Interface Functions on page 151.
Solaris 8 Release
The secure directory from which files can be preloaded is now /usr/lib/secure
for 32bit objects and /usr/lib/secure/64 for 64bit objects. See Security
on page 80.
Greater flexibility in modifying the runtime linkers search paths can be achieved
with the link-editors -z nodefaultlib option, and runtime configuration files
created by the new utility crle(1). See Directories Searched by the Runtime
Linker on page 35 and Configuring the Default Search Paths on page 67.
The new EXTERN mapfile directive enables you to use -z defs with externally
defined symbols. See Defining Additional Symbols on page 46.
The new $ISALIST, $OSNAME, and $OSREL dynamic string tokens provide greater
flexibility in establishing instruction set specific, and system specific dependencies.
See Dynamic String Tokens on page 67.
The link-editor options -p and -P provide additional means of invoking runtime
link auditing libraries. See Recording Local Auditors on page 157. The runtime
link auditing interfaces la_activity() and la_objsearch() have been added.
Solaris 7 Release
The 64bit ELF object format is now supported. See File Format on page 177 for
details. Link-editor extensions and differences for 64bit processing include the use
of /usr/lib/64 (see Directories Searched by the Link-Editor on page 33,
Directories Searched by the Runtime Linker on page 35, and Naming
Conventions on page 102), the environment variable LD_LIBRARY_PATH_64 (see
Using an Environment Variable on page 34, and Directories Searched by the
Runtime Linker on page 64), and the runtime linker /usr/lib/64/ld.so.1
(see Chapter 3).
You can build shared objects with optimized relocation sections using the
link-editors -z combreloc option. See Combined Relocation Sections
on page 121.
The new $ORIGIN dynamic string token provides greater flexibility in establishing
dependencies within unbundled software. See Dynamic String Tokens
on page 67.
The loading of a shared object can now be deferred until the object is actually
referenced by the running program. See Lazy Loading of Dynamic Dependencies
on page 114.
The new SHT_SUNW_COMDAT section type enables the elimination of
multiply-defined symbols. See COMDAT Section on page 203.
The new SHT_SUNW_move section type enables partially initialized symbols. See
Move Section on page 206.
The runtime link auditing interfaces la_symbind64(), la_sparcv9_pltenter
(), and la_pltexit64(), together with a new link-auditing flag
LA_SYMB_ALTVALUE, have been added. See Audit Interface Functions
on page 158.
335
D dynamic information tags
data representation, 179 NEEDED, 65, 103
debugging aids RUNPATH, 65
link-editing, 60 SONAME, 103
runtime linking, 93 SYMBOLIC, 125
demonstrations TEXTREL, 115
prefcnt, 163 dynamic linking, 22
sotruss, 163 implementation, 210, 244
symbindrep, 163
whocalls, 163
dependency
groups, 83, 84 E
dependency ordering, 106 ELF, 19, 25, 102, 112, 149
direct binding, 69, 70, 121 See also object files
dladdr(3DL), 331 elf(3E), 23, 149
dladdr1(3DL), 331 environment variables, 23
dlclose(3DL), 77, 81 LD_AUDIT, 80, 157
dldump(3DL), 37 LD_BIND_NOT, 95
dlerror(3DL), 81 LD_BIND_NOW, 72, 79, 95, 264, 267, 269
dlfcn.h, 81 LD_BREADTH, 78
dlinfo(3DL), 330, 331, 332 LD_CONFIG, 80
dlmopen(3DL), 156 LD_DEBUG, 93
See also dlopen(3DL) LD_DEBUG_OUTPUT, 94
dlopen(3DL), 64, 81, 82, 88, 111, 139, 149, 150 LD_LIBRARY_PATH, 34, 66, 80, 82, 106, 157
effects of ordering, 87 LD_LOADFLTR, 111
group, 84 LD_NOAUDIT, 158
dlopen(3DL) LD_NOAUXFLTR, 111
group, 83 LD_NODIRECT, 70
dlopen(3DL) LD_NOLAZYLOAD, 76
modes LD_NOVERSION, 139, 142
RTLD_GLOBAL, 83, 88 LD_OPTIONS, 27, 61
RTLD_GROUP, 88 LD_PRELOAD, 70, 73, 80
RTLD_LAZY, 83 LD_PROFILE, 125
RTLD_NOLOAD, 156 LD_PROFILE_OUTPUT, 126
RTLD_NOW, 72, 79, 83 LD_RUN_PATH, 36
RTLD_PARENT, 88, 89 LD_SIGNAL, 80
of a dynamic executable, 88 SGS_SUPPORT, 150
dlopen(3DL) error messages
of a dynamic executable, 83 link-editor
dlopen(3DL) illegal argument to option, 28
shared object naming conventions, 102 illegal option, 28
dlsym(3DL), 64, 81, 89, 92, 140, 150 incompatible options, 28
special handle multiple instances of an option, 28
RTLD_DEFAULT, 45, 89 multiply-defined symbols, 42
RTLD_NEXT, 89 relocations against non-writable
RTLD_SELF, 332 sections, 115
dump(1), 23, 65, 68, 113, 115 shared object name conflicts, 105
dynamic executables, 20, 21 soname conflicts, 105
Index 337
link-editing (Continued) link-editor options (Continued)
input file processing, 28 -t, 41, 42
library input processing, 29 -u, 46, 47
library linking options, 29 -Y, 34
mixing shared objects and archives, 32 -z allextract, 29
position of files on command line, 32 -z combreloc, 309
search paths, 33 -z defaultextract, 29
shared object processing, 30 -z defs, 44, 48, 156, 309
link-editor, 19, 25 -z direct, 329
debugging aids, 60 -z endfiltee, 259
error messages -z finiarray, 36
See error messages -z groupperm, 260
external bindings, 56 -z ignore, 30, 117, 309
invoking directly, 26 -z initarray, 36
invoking using compiler driver, 27 -z initfirst, 258
overview, 25 -z interpose, 70, 259
sections, 25 -z lazyload, 75, 260, 309, 310
segments, 25 -z ld32, 150
specifying options, 27 -z ld64, 150
link-editor options -z loadfltr, 111, 258
-64, 23, 109 -z muldefs, 42
-a, 308 -z nocompstrtab, 56, 330
-B direct, 309, 310 -z nodefaultlib, 35, 259
-B dynamic, 32 -z nodefs, 43, 72
-B eliminate, 56 -z nodelete, 258
-B group, 84, 88, 258 -z nodirect, 329
-B local, 54 -z nodlopen, 258
-B reduce, 49, 55 -z nodump, 259
-B static, 32, 308 -z nolazyload, 75
-D, 60 -z nopartial, 208
-d n, 307, 310 -z noversion, 54, 132, 138
-d y, 308 -z now, 72, 79, 83
-e, 58 -z record, 118
-F, 107 -z rescan, 33
-f, 107 -z text, 115, 308
-G, 101, 308, 310 -z verbose, 59
-h, 65, 103, 147, 309 -z weakextract, 29, 225
-i, 34 link-editor output
-L, 33, 307 dynamic executables, 20
-l, 29, 31, 102, 146, 307 relocatable objects, 20
-M, 26, 46, 47, 130, 131, 140, 293, 309, 313 shared objects, 20
-m, 31, 40 static executables, 20
-P, 157 link-editor support interface (ld-support), 149
-p, 157 ld_atexit(), 153
-R, 35, 105, 309, 310 ld_atexit64(), 153
-r, 27, 308 ld_file(), 151
-S, 150 ld_file64(), 151
-s, 55, 57 ld_input_done(), 153
Index 339
procedure linkage table (Continued) runtime linker (Continued)
SPARC, 217, 218, 261, 264 shared object processing, 64
x86, 220, 221, 267 version definition verification, 138
profil(2), 126 runtime linker support interfaces
program interpreter, 247 (rtld-audit), 149, 155
See also runtime linker la_activity(), 159
pvs(1), 23, 132, 134, 136, 137 la_i86_pltenter(), 161
la_objclose(), 162
la_objopen(), 159
la_objseach(), 159
R la_pltexit(), 162
relocatable objects, 20 la_preinit(), 160
relocation, 67, 120, 124, 210 la_sparcv8_pltenter(), 161
copy, 58, 122 la_sparcv9_pltenter(), 161
displacement, 58 la_symbind32(), 160
immediate, 71 la_symbind64(), 160
lazy, 71 la_version(), 158
non-symbolic, 68, 120 runtime linker support interfaces
runtime linker (rtld-debugger), 149, 164
symbol lookup, 68, 71, 83, 95 ps_global_sym(), 175
symbolic, 68, 120 ps_pglobal_sym(), 176
RTLD_DEFAULT, 45 ps_plog(), 175
See also dependency ordering ps_pread(), 175
RTLD_GLOBAL, 83, 88 ps_pwrite(), 175
RTLD_GROUP, 88 rd_delete(), 168
RTLD_LAZY, 83 rd_errstr(), 168
RTLD_NEXT, 89 rd_event_addr(), 171
RTLD_NOLOAD, 156 rd_event_enable(), 171
RTLD_NOW, 72, 79, 83 rd_event_getmsg(), 172
RTLD_PARENT, 88, 89 rd_init(), 167
runpath, 35, 65, 82, 105 rd_loadobj_iter(), 170
RUNPATH, See runpath rd_log(), 168
runpath, security, 80 rd_new(), 167
runtime environment, 21, 31, 101 rd_objpad_enable(), 174
runtime linker, 21, 63, 247 rd_plt_resolution(), 173
direct binding, 69, 70, 121 rd_reset(), 167
initialization and termination routines, 76 runtime linking, 21
lazy binding, 71, 83, 95
link-maps, 156
loading additional objects, 73
namespace, 156 S
programming interface SCD, See Application Binary Interface
See also dlclose(3DL), dldump(3DL), search paths
dlerror(3DL), dlmopen(3DL), and link-editing, 33
dlopen(3DL) runtime linker, 35, 64
relocation processing, 67 $ISALIST token, 321
search paths, 35, 64 $ORIGIN token, 323
security, 80 $OSNAME token, 323
Index 341
section types (Continued) symbol reserved names (Continued)
SHT_SYMTAB_SHNDX, 193 _PROCEDURE_LINKAGE_TABLE_, 58
sections, 25, 112 _start, 58
See also section flags, section names, section _START_, 58
numbers and section types symbol resolution, 38, 57
security, 80, 326 complex, 41
segments, 25, 112 fatal, 42
data, 113, 114 interposition, 69
text, 113, 114 multiple definitions, 30
SGS_SUPPORT, 150 search scope
shared libraries, See shared objects group, 84
shared objects, 19, 20, 21, 64, 101 simple, 39
as filters, 107 symbol visibility, 224, 226
dependency ordering, 106 global, 84
explicit definition, 44 local, 84
implementation, 210, 244 SYMBOLIC, 125
implicit definition, 44 symbols
link-editor processing, 30 absolute, 48, 188
naming conventions, 31, 102 archive extraction, 29
recording a runtime name, 103 auto-elimination, 55
with dependencies, 105 auto-reduction, 48, 132, 314
size(1), 112 COMMON, 38, 48, 51, 188
Solaris ABI, See Application Binary Interface defined, 38
Solaris Application Binary Interface, See definition, 29, 42
Application Binary Interface elimination, 55
SONAME, 103 existence test, 45
SPARC Compliance Definition, See Application global, 38, 39, 129, 224, 227
Binary Interface local, 38, 224, 227
standard filters, 107, 108 multiply-defined, 30, 40, 203
static executables, 20 ordered, 188
strings(1), 119 private interface, 129
strip(1), 55, 57 public interface, 129
SUNWosdem, 163, 166, 177 reference, 29, 42
SUNWtoo, 163 registers, 229
support interfaces runtime lookup, 84, 92
link-editor (ld-support), 149 deferred, 71, 83, 95
runtime linker (rtld-audit), 149, 155 scope, 84, 88
runtime linker (rtld-debugger), 149, 164 symbol visibility, 84
symbol reserved names, 57 tentative, 29, 38, 46, 48, 51, 188
_DYNAMIC, 57 ordering in the output file, 46
_edata, 57 realignment, 51
_end, 57 type, 225
_END_, 57 undefined, 29, 38, 42, 43, 44, 188
_etext, 57 visibility, 224, 226
_fini, 36 weak, 29, 39, 45, 224, 227
_GLOBAL_OFFSET_TABLE_, 58, 116, 261 System V Application Binary Interface, 312
_init, 36
main, 58
T
tentative symbols, 29, 38, 48, 51
TEXTREL, 115
Thread Local Storage, 330
thread-local storage, 271
section definition, 272
TLS, See thread-local storage
tsort(1), 30, 61
U
undefined symbols, 42
/usr/ccs/bin/ld, See link-editor
/usr/ccs/lib, 33
/usr/lib, 33, 35, 64, 65, 82
/usr/lib/64, 33, 35, 64, 65, 82
/usr/lib/64/ld.so.1, 63, 164
/usr/lib/ld.so.1, 63, 164
/usr/lib/secure, 80, 157
/usr/lib/secure/64, 80, 157
V
versioning, 129
base version definition, 132
binding to a definition, 136, 140
$ADDVERS, 140
defining a public interface, 54, 131
definitions, 130, 131, 136
file control directive, 140
file name, 131, 315
generating definitions within an image, 47,
54, 131
normalization, 138
overview, 129
runtime verification, 138, 139
virtual addressing, 241
Index 343
344 Linker and Libraries Guide April 2004